Incidence and predictors of lost to follow-up in “test and treat era” among adults on ART in Eastern Ethiopia: a retrospective follow-up study

Abstract

Background

Despite expansion of ART under universal test-and-treat strategies, lost to follow-up remains a major barrier to effective HIV care, leading to treatment interruption, drug resistance, and poor clinical outcomes. Evidence on lost to follow-up in Eastern Ethiopia during the test-and-treat era remains limited. Therefore, this study aimed to assess the incidence and predictors of lost to follow-up in “test and treat era” among adults on Anti-Retroviral Treatment at Public Hospitals in Eastern Ethiopia.

Method

A retrospective follow-up study was conducted among 863 randomly selected adults newly enrolled in ART after the implementation of the Universal test and treat strategy from November 15, 2016 to March 30, 2024. LTFU was defined as failure to return for ART refill for ≥90 days after the last missed appointment without documented death, transfer-out, or treatment discontinuation. Data were extracted from patient charts using koboTool, and analyzed using STATA (V17.0). Kaplan–Meier survival curve with log-rank test used to compare survival differences, and Cox regression was used to identify predictors of lost to follow-up. Proportional hazard assumptions were checked.

Results

The median follow-up time was 2.1 years. A total of 229 study participants (26.54%) were lost to follow-up, yielding an incidence rate of 9.9 (95% CI:8.72–11.30) per 100 person-years. Higher hazard of lost to follow-up was observed among those without a contact person (AHR = 1.47;95% CI:1.01–2.12), who had not disclosed HIV status (AHR = 1.51;95% CI:1.03–2.20), presented with WHO clinical stage IV (AHR = 2.86;95% CI:1.57–5.18), baseline CD4 < 200 cells/mm³ (AHR = 1.61;95% CI:1.03–2.51), opportunistic infection (AHR = 1.44;95% CI:1.02–2.04), no isoniazid preventive therapy (AHR = 1.60;95% CI:1.18–2.18) and poor adherence (AHR = 2.51;95% CI:1.81–3.47). Conversely, BMI < 18.5 kg/m² was linked to lower hazard of LTFU (AHR = 0.65;95% CI:0.45–0.92).

Conclusion

LTFU was relatively high and remains a substantial challenge among adults receiving ART in Eastern Ethiopia during the UTT era. Clinical severity, poor adherence, and limited social support were important predictors of LTFU. Strengthening patient monitoring, adherence counseling, and targeted support for high-risk patients is essential to improve retention in HIV care and progress toward the UNAIDS 95–95-95 targets.

Background

HIV/AIDS remains a major global health problem having claimed about 40.4 million lives worldwide [1, 2]. The expansion of antiretroviral therapy (ART) has substantially reduced HIV related morbidity and mortality contributing to 38% decline in new HIV infections and a 52% decline in AIDS-related deaths [1, 3]. To further accelerate HIV epidemic control, the Joint United Nations Program on HIV/AIDS (UNAIDS) introduced the 95–95-95 targets, while the World Health Organization (WHO) recommended the Universal Test and Treat (UTT) strategy in 2015, promoting rapid initiation of antiretroviral therapy, including same-day or within seven days of HIV diagnosis, regardless of CD4 count or clinical stage [4–6].

Ethiopia adopted UTT strategy in 2016, resulting in considerable increase in ART access and coverage [7, 8]. Despite these achievements lost to follow-up (LTFU) remain a major challenge to successful HIV acre. LTFU refers to patients who fail to return for HIV care within a specified period without documented death, transfer-out, or treatment discontinuation [9, 10].

Evidence indicated that the burden of LTFU remains substantial in low- and middle-income countries, particularly among people living with HIV receiving long-term ART care [11]. In Ethiopia, LTFU range from 3.7 to 26.6 per 100 person-years with variation across populations and settings [12–17].

Several factors affect LTFU among people living with HIV. Socio-demographic factors such as lack of social support and non-disclosure of HIV status may reduce retention in care, while advanced disease stage, opportunistic infections, and poor treatment adherence can increase the likelihood of disengagement from treatment services [18–22]. In some contexts higher LTFU were reported among individuals initiated on ART on the same [23].

The consequences of LTFU are substantial, as it compromise treatment success, increases HIV transmission risk and drug resistance, and ultimately contributes to preventable morbidity and mortality [18, 24]. In resource limited settings such as Ethiopia, rising LTFU has also been associated with continued HIV transmission with evidence suggesting up to 61% of new HIV transmissions may originate from individuals lost to care [25, 26]. Therefore, generating updated evidence on LTFU is essential for improving ART retention and supporting progress toward the UNAIDS 95–95-95 targets.

Despite availability of previous studies that have investigated LTFU among people living with HIV most were conducted during early phase of UTT era and evidence from Eastern Ethiopia remains scarce. In addition, this study considered important variables such as distance from health facilities and travel time, which are not addressed in previous Ethiopian studies on time to LTFU. Therefore, this study aimed to determine the incidence and predictors of LTFU among people living with HIV receiving ART in the “test and treat” era in Eastern Ethiopia, providing updated context-specific evidence to support interventions aimed at improving retention and progress toward the UNAIDS 95–95-95 targets.

Methods

Study design, setting and period

An institutional-based retrospective follow-up study was conducted in public hospitals of Harari Region and Dire Dawa City Administration, Eastern Ethiopia. Data were collected from patient charts between September 2 and November 2, 2024 and patients enrolled in ART were followed from November 15, 2016, to March 30, 2024. The study period was selected to capture outcomes after Ethiopia implemented the Universal Test and Treat (UTT) strategy in 2016.

Harari Region has an estimated population of 283,000 [27]. HIV care at Hospital level is mainly provided by two public hospitals, both of which were included in this study: Hiwot Fana Comprehensive Specialized University Hospital (HFCSUH) and Jugal General Hospital. HFCSUH is a major referral and teaching hospital serving about 5.8 million people in Eastern Ethiopia and has provided ART service since 1995. Since the implementation of UTT strategy 598 HIV-positive adults were newly initiated on ART. Jugal General Hospital is also the region’s main HIV treatment center since 1996, serves more than 1 million people, with 1,165 HIV-positive adults enrolled in ART follow-up since UTT implementation.

Dire Dawa city administration, one of the two chartered city administrative in Ethiopia, has an estimated population of 551,000 [27]. There are two public hospitals, but only Dilchora Referral Hospital was included in this study, as the second hospital was a health center until recently upgraded to hospital status. Dilchora Referral Hospital serves about 1.5 million people, has provided ART service since 1996, and enrolled 858 HIV-positive adults on ART follow-up since UTT implementation. Inclusion of hospitals with different patient loads and service capacities improves the representativeness and generalizability of the findings.

Sample size determination, sampling procedure and participants

The sample size was determined using STATA version 17.0. For the first objective, the sample size was estimated using the log-rank survival comparison method, while the “stpower cox” command was applied for predictors of LTFU. Assumptions included 95% confidence level, 80% power, 10% withdrawal rate, and hazard ratios/probabilities of events obtained from previous studies on predictors including WHO clinical stage [28], age [15], residence [12] and cotrimoxazole preventive therapy [28]. These variables were selected because they were consistently associated with LTFU in prior studies. The final sample size was 912.

All HIV-positive adults newly enrolled in ART between November 15, 2016, and March 30, 2024, at the three public hospitals following the UTT strategy were eligible for the study. The sample was proportionally allocated to each hospital, and participants were selected by simple random sampling using computer-generated numbers (Fig. 1).

Fig. 1.

Schematic presentation of the sampling procedure for the study to be conducted among PLHIV on ART at public hospitals in Eastern Ethiopia, 2024. HFCSUH is Hiwot Fana Comprehensive Specialized University Hospital, N = is the total number of HIV positive adults who were newly enrolled in ART from November 15, 2016 to March 30, 2024 at each Hospital n = is the number card of HIV positive adult patients from which data was extracted in each hospital nt = is a total sample size

Individuals aged 15 and above years and with at least one active follow-up visit were included, while those transferred in from other facilities after ART initiation were excluded from the study.

Variables

The dependent variable was incidence of LTFU. Independent variables included socio-demographic factors (Age, Sex, Religion, Marital status, Educational status, Occupation, Residence, Having registered cell phone, having caregiver, person to contact, HIV status disclosure, distance from the facility, time taken to reach the facility, family size), Clinical factors (Partner’s HIV status, CD4 cell count, viral load count, WHO clinical stage, BMI, Opportunistic infections, Functional status, TB/HIV co-infection, comorbidity) and treatment related factors (ART regimen type, ART initiation time, ART drug change, ART regimen change, ART drug adherence, INH prophylaxis provision, Cotrimoxazole preventive therapy, Presence of side effect).

Operational definitions

LTFU was defined as failure to attend ART clinics for more than 90 days after the last appointment date without documentation of death or formal transfer-out based on national and previous study definitions [29, 30]. Patients who remained active on ART, transferred out, died, or completed follow-up without experiencing LTFU were treated as censored observations [31]. Time to LTFU was the interval between ART initiation and occurrence of LTFU or censoring [31].

ART initiation time referred to the interval between HIV diagnosis and ART initiation and was categorized as rapid initiation (within 7 days) and delayed initiation (after 7 days), consistent with WHO UTT recommendations [5, 32].

Transferred-out patients were those formally moved to another facility with documented transfer [14]. ART drug adherence was categorized as good (≥95%), fair (85–94%), or poor (<85%) [29, 33] based on the proportion of prescribed doses taken, following national HIV treatment monitoring guidelines. Functional status was classified as working (able to perform usual work), ambulatory (able to perform daily activities but unable to work), or bedridden (unable to perform daily activities) according to national ART guidelines [29].

Data collection tool and procedure

Data were extracted from ART medical charts and electronic databases using a structured checklist adapted from the standardized Federal Ministry of Health (FMOH) ART intake and follow-up forms and relevant literatures [15, 28, 31, 33–36]. Three BSc health professionals with experience in HIV care collected the data, and two supervisors were assigned to closely monitor the data collection process.

Data quality

Training was provided for both data collectors and supervisors before data collection. The data extraction tool was reviewed by senior experts to the area of study to ensure content validity. A pretest was conducted on 5% of the sample size one week prior to data collection to assess the consistency of facilities documentation with the extraction format and its alignment with the study objectives. Necessary adjustments were made to the tool based on the pretest findings and experiences.

Data processing and analysis

Data were exported from koboTool to Microsoft Excel and then imported into STATA version 17.0 for statistical analysis. Data cleaning was performed to check for inconsistencies and missing values before statistical analysis. Continuous variables were summarized using median and interquartile range (IQR), while categorical variables were presented using frequencies and percentages. Tables and graphs were used to present the findings.

The incidence rate of LTFU was calculated by dividing the number of patients who experienced LTFU during the follow-up period by the total person-years of observation. Person-time was calculated from ART initiation to the date of LTFU or censoring. Kaplan–Meier survival curve, along with the log-rank test were used compare survival across categories of covariates.

A Cox proportional hazards regression was applied to identify predictors of time to LTFU. Variables with p ≤ 0.25 in bivariable analysis were included in multivariable model to avoid excluding potentially important predictors. Multivariable Cox regression was then performed to identify independent predictors of time to LTFU. The proportional hazard assumption was verified using log-log survival probability plot and Schoenfeld residual test. Multicollinearity was assessed through variance inflation factor (VIF). The goodness of fit of final model was also evaluated using the Cox–Snell residual plot. Ties in survival time were handled using the Efron approximation method. Possible interaction terms among key variables were assessed during model building, but no significant interaction was identified. Finally p-value < 0.05 was considered statistically significant, and adjusted hazard ratio (AHR) with 95% confidence interval (CI) were reported to indicate the strength and direction of associations.

Results

Socio-demographic characteristics

Of the 912 reviewed records, 49 were excluded due to incomplete medical information (missing ART initiation date and undefined outcome), and 863 records were included in the final analysis. The median age of the participants was 35 years (IQR: 28–40). The majority were female (524; 60.72%) and married (357; 41.37%). Most participants (736, or 85.28%) resided in urban areas. Only 163 participants (18.89%) had attained tertiary educational, while 197 (22.83%) were unemployed. More than half (448; 51.91%) had disclosed their HIV status, and nearly three-quarters (637; 73.81%) had a cell phone to contact (Table 1).

Table 1.

Socio-demographic characteristics of HIV-positive adults on ART by follow-up outcome at public hospitals in Eastern Ethiopia between November 2016 and March 2024 (N = 863)

Variables	Category	Outcome status		Frequency	Percentage (%)
		Event (n = 229)	Censored (n = 634)
Enrollment age	15–24	53	88	141	16.34
	25–34	78	190	268	31.05
	35–44	71	216	287	33.26
	≥45	27	140	167	19.35
Sex	Male	81	258	339	39.28
	Female	148	376	524	60.72
Religion	Orthodox	88	325	413	47.86
	Muslim	108	230	338	39.17
	Protestant	31	72	103	11.93
	Other	2	7	9	1.04
Marital status	Single	53	102	155	17.96
	Married	77	280	357	41.37
	Divorced	72	151	223	25.84
	Widowed	27	101	128	14.83
Educational status	No formal education	64	123	187	21.67
	Primary education	75	211	286	33.14
	Secondary education	53	174	227	26.30
	Tertiary education	37	126	163	18.89
Occupation	Unemployed/Jobless	65	132	197	22.83
	Governmental/NGO employed	29	130	159	18.42
	Daily labor	47	127	174	20.16
	Student	18	32	50	5.79
	Housewife	35	79	114	13.21
	Merchant	23	92	115	13.33
	Farmer	6	17	23	2.67
	Other	6	25	31	3.59
Residence	Urban	170	566	736	85.28
	Rural	59	68	127	14.72
Distance from the facility	≤20 km	163	513	676	78.33
	>20 km	66	121	187	21.67
Time to reach the facility from home	<1 hr	170	524	694	80.42
	1-2 hr	14	41	55	6.37
	>2 hr	45	69	114	13.21
Having registered cell phone	Yes	142	495	637	73.81
	No	87	139	226	26.19
Having Care giver	Yes	102	480	582	67.44
	No	127	154	281	32.56
Had a person to contact	Yes	76	397	473	54.81
	No	153	237	390	45.19
Family size	No family member	42	81	123	14.25
	1–3 family member	148	438	586	67.90
	≥4 family member	39	115	154	17.84
Disclosure status	Disclosed	71	377	448	51.91
	Not disclosed	158	257	415	48.09

Religion other includes- Catholic, Jova, Occupation other includes- Sex-worker, Driver

Clinical characteristics

At ART initiation, 505 participants (58.52%) had working functional status, and 518 (60.02%) were in WHO clinical stage I. The median baseline CD4 cell count was 413 cell/mm³ (IQR: 219–680), with 362 participants (41.95%) having a baseline CD4 ≥ 500 cell/mm³. The median baseline BMI was 20.76 kg/m² (IQR: 18.18–24). Nearly half of the participants (49.13%) had an opportunistic infection, most commonly tuberculosis (58.49%). Comorbidities were present in 151 participants (17.50%) (Table 2).

Table 2.

Clinical characteristics of HIV-positive adults on ART by follow-up outcome at public hospitals in Eastern Ethiopia between November 2016 and March 2024 (N = 863)

Variables	Category	Outcome status		Frequency	Percentage (%)
		Event (n = 229)	Censored (n = 634)
Partners HIV status	Positive	29	138	167	25.57
	Negative	8	66	74	11.33
	Unknown	129	283	412	63.10
Baseline functional status	Working	101	404	505	58.52
	Ambulatory	92	182	274	31.75
	Bedridden	36	48	84	9.73
OI	Yes	154	270	424	49.13
	No	75	364	439	50.87
Type of OI n = 154	Zoster	28	23	51	12.03
	Candidiasis	52	70	122	28.77
	Pneumonia	45	66	111	26.18
	TB	94	154	248	58.49
	Other	39	87	126	29.72
Last known TB status	Positive	44	66	110	12.75
	Negative	185	568	753	87.25
Comorbidity other than OI	Yes	47	104	151	17.50
	No	182	530	712	82.50
Types of comorbidity n = 47	Vascular diseases	16	20	36	23.84
	Hypertension	8	12	20	13.25
	Diabetes Mellitus	6	17	23	15.23
	Kidney diseases	7	25	32	21.19
	Other	10	30	40	26.49
Baseline WHO clinical stage	Stage I	98	420	518	60.02
	Stage II	51	104	155	17.96
	Stage III	57	81	138	15.99
	Stage IV	23	29	52	6.03
Baseline BMI	<18.5 kg/m2	59	178	237	27.46
	18.5–24.9 kg/m2	116	346	462	53.54
	≥25 kg/m2	54	110	164	19.00
Baseline CD4 count	<200 mm3	58	123	181	20.97
	200–499 mm3	113	207	320	37.08
	≥500 mm3	58	304	362	41.95
Baseline viral load	≤1000 copies/ml	177	590	767	88.88
	>1000 copies/ml	44	52	96	11.12

OI- opportunistic infection, BMI-body mass index, Other OI includes- CNS toxoplasmosis, cervical cancer, cryptococcal meningitis, Diarrhea acute and chronic, extra-pulmonary TB, oral thrush, other comorbidity includes- Covid-19, liver disease, spleen disease

Treatment related characteristics

More than three-quarters of participants (695; 80.53%) initiated ART within 7 days of confirmation of diagnosis, of whom 543 (62.92%) started on the same day as diagnosis. Over half (514; 59.56%) began treatment with TDF-3TC-DTG, followed by 321 (37.20%) on TDF/3TC/EFV. Drug changes were reported in 251 participants (29.08%). Cotrimoxazole Preventive Therapy and Isoniazid Preventive Therapy were provided to 577 (66.86%) and 522 (60.49%) participants, respectively. At baseline, 581 participants (67.32%) demonstrated good ART drug adherence, while only 89 (10.31%) reported experiencing drug side effects (Table 3).

Table 3.

Treatment related characteristics of HIV-positive adults on ART by follow-up outcome at public hospitals in Eastern Ethiopia between November 2016 and March 2024 (N = 863)

Variables	Category	Outcome status		Frequency	Percentage (%)
		Event (n = 229)	Censored (n = 634)
Overall ART initiation time	Within 07 days	176	519	695	80.53
	After 07 days	53	115	168	19.47
Baseline ART regimen	1j	127	387	514	59.56
	1c	3	5	8	0.93
	1d	1	2	3	0.35
	1e	95	226	321	37.19
	1f	1	7	8	0.93
	Other	2	7	9	1.04
Experience of any drug change	Yes	64	187	251	29.08
	No	165	447	612	70.92
Experience of drug regimen change	Yes	1	7	8	0.93
	No	228	627	855	99.07
IPT status	Received	98	424	522	60.49
	Not received	131	210	341	39.51
CPT status	Received	115	462	577	66.86
	Not received	114	172	286	33.14
History of drug side effect	Yes	28	61	89	10.31
	No	201	573	774	89.69
Baseline ART drug adherence level	Good	92	489	581	67.32
	Fair	51	59	110	12.75
	Poor	86	86	172	19.93

IPT- Isoniazid preventive therapy, CPT- Cotrimoxazole preventive therapy, 1j- TDF+3TC+DTG, 1c-AZT+3TC+NVP, 1d-AZT+3TC+EFV, 1e-TDF+3TC+EFV, 1f-TDF+3TC+NVP, Other ART regimen includes- 2f- AZT+3TC+ATV/r, 2 h- TDF+3TC+ATV/r, 2i- ABC+3TC+LPV/r

Incidence of LTFU

Participants’ median follow up time was 2.1 years with minimum of 7 days to a maximum of 8.5 years, contributing a total of 2296.4 person-years of observation. The median survival time was undefined, indicating that the study ends before 50% of participants experienced LTFU. Upon the end of the follow-up, 229 participants (26.54%, 95% CI: 23.69–29.59) had been LTFU, while more than half 463 (53.65%, 95% CI: 50.31–56.96) remained active, 99 (11.47%, 95% CI: 9.51–13.78) had died, and 72 (8.34%, 95% CI: 6.67–10.39) had transferred out. The overall incidence rate of LTFU was 9.9 per 100 person-years of observation (95% CI: 8.72–11.30).

Kaplan–Meier failure estimate

The Kaplan–Meier failure curve demonstrated that the likelihood of LTFU increased over the follow-up period (Fig. 2). Cumulative hazard estimate of LTFU varied across categories of different independent variables. Participants with a baseline CD4 count < 200 cell/mm³ had the highest cumulative hazard of LTFU, followed by those with 200–499 cell/mm³, while those with ≥500 cell/mm³ had the lowest risk (log rank, p < 0.001) (Fig. 3).

Fig. 2.

Kaplan-Meier failure estimate of the overall cumulative probability of lost to follow-up among HIV-positive adults on ART at public hospitals in Eastern Ethiopia between November 2016 and March 2024

Fig. 3.

Kaplan-Meier failure estimate of lost to follow-up by their baseline CD4 cell count at public hospitals in Eastern Ethiopia between 2024

Similarly, HIV status disclosure, presence of opportunistic infection, Isoniazid preventive therapy status and having a person to contact, also showed marginally significant differences across their categories (Fig. 4).

Fig. 4.

Kaplan-Meier failure estimate of lost to follow-up using disclosure status (A), opportunistic infection (B), Isoniazid preventive therapy status (C) and person to contact (D) At public hospitals in Eastern Ethiopia between 2024

Predictors of lost to follow-up

In bivariable cox regression, eighteen variables met the inclusion criterion (p ≤ 0.25) and were entered into the multivariable model. In the final multivariable Cox regression analysis, eight factors, namely having person to contact, disclosure status, opportunistic infection, baseline WHO clinical stage, baseline BMI, baseline CD4 cell count, isoniazid preventive therapy status and baseline drug adherence, were identified as the major predictor of LTFU at p < 0.05. The proportional hazard assumption was satisfied (Schoenfeld test p = 0.482; log-log plots), and model fit was confirmed by the Cox–Snell residual plot showing close alignment of the Nelson-aelon curve with the 45-degree reference line (Fig. 5).

Fig. 5.

Cox-snell test graph showing the overall model goodness of fit result for incidence of lost to follow up and its predictors among HIV-positive adults on ART at public hospitals in Eastern Ethiopia, 2024

The hazard of LTFU among participants who had no person to contact was 1.47 times higher as compared to those who had person to contact at any given time (AHR = 1.47; 95% CI: 1.01–2.12). Furthermore the risk of LTFU among participants who did not disclose their HIV status was 1.51 (AHR = 1.51; 95% CI: 1.03–2.20) times higher than among those who disclosed their HIV status at any given time during the study period.

Compared with WHO clinical stage I, the hazard of LTFU was twice as high in stage III (AHR = 2.09; 95% CI: 1.41–3.12) and nearly three times higher in stage IV (AHR = 2.86; 95% CI: 1.57–5.18). Likewise, participants with baseline CD4 cell counts < 200 cells/mm³ and 200–499 cell/mm³ had significantly higher hazards than those with ≥500 cells/mm³ (AHR = 1.61; 95% CI: 1.03–2.51 and AHR = 1.54; 95% CI: 1.07–2.22, respectively).

Moreover participants who developed opportunistic infection were 1.44 times more likely to be LTFU as compared to those who did not develop any opportunistic infection throughout study period (AHR = 1.44; 95% CI: 1.02–2.04). Conversely, participants with baseline BMI < 18.5 kg/m² had a 35% lower hazard of LTFU compared with those with normal BMI (18.5–24.9 kg/m²) (AHR = 0.65; 95% CI: 0.45–0.92), possibly reflecting closer clinical monitoring and nutritional support among undernourished patients.

HIV-positive adults who did not receive IPT were 1.6 times at higher hazard of being lost from their follow-up when compared to those who did receive IPT (AHR = 1.60; 95% CI: 1.18–2.18) throughout study period. Furthermore, suboptimal ART adherence markedly increased the risk, with fair adherence nearly tripling the hazard (AHR = 2.80; 95% CI: 1.92–4.09) and poor adherence more than doubling it (AHR = 2.51; 95% CI: 1.81–3.47) compared to good adherence (Table 4).

Table 4.

Bivariable and multivariable Cox regression analysis of predictors of LTFU among HIV-positive adults on ART at public hospitals in Eastern Ethiopia, 2024

Variables	Category	Outcome status		CHR (95% CI)	AHR (95% CI)	p-value
		Event (n = 229)	Censored (n = 634)
Sex	Male	81	258	0.82 (0.63–1.08)	0.92 (0.65–1.29)	0.622
	Female	148	376	1	1
Enrollment age	15–24	53	88	1	1
	25–34	78	190	0.70 (0.49–0.99)	0.76 (0.49–1.17)	0.217
	35–44	71	216	0.63 (0.44–0.89)	0.77 (0.48–1.22)	0.263
	≥45	27	140	0.48 (0.30–0.77)	0.54 (0.29–1.01)	0.053
Marital status	Single	53	102	1.76 (1.24–2.50)	1.19 (0.76–1.87)	0.439
	Married	77	280	1	1
	Divorced	72	151	1.70 (1.23–2.35)	1.22 (0.83–1.78)	0.311
	Widowed	27	101	1.21 (0.78–1.87)	0.97 (0.58–1.62)	0.918
Educational status	No formal education	64	123	1.93 (1.29–2.90)	0.91 (0.49–1.70)	0.777
	Primary education	75	211	1.29 (0.87–1.92)	0.88 (0.49–1.59)	0.683
	Secondary education	53	174	1.04 (0.68–1.58)	0.79 (0.45–1.37)	0.399
	Tertiary education	37	126	1	1
Occupation	Unemployed/Jobless	65	132	1.38 (0.95–2.01)	1.24 (0.81–1.90)	0.321
	Governmental/NGO employed	29	130	0.57 (0.36–0.91)	0.67 (0.35–1.29)	0.238
	Daily labor	47	127	1	1
	Student	18	32	1.36 (0.79–2.34)	1.43 (0.69–2.98)	0.338
	Housewife	35	79	1.25 (0.80–1.93)	1.51 (0.87–2.59)	0.137
	Merchant	23	92	0.60 (0.36–1.00)	0.68 (0.39–1.17)	0.168
	Farmer	6	17	1.24 (0.53–2.89)	0.76 (0.30–1.97)	0.583
	Other	6	25	0.71 (0.30–1.66)	0.90 (0.38–2.17)	0.824
Distance from the facility	≤20 km	163	513	1	1
	>20 km	66	121	1.51 (1.13–2.01)	1.01 (0.45–2.24)	0.984
Time to reach the facility from home	<1 hr	170	524	1	1
	1-2 hr	14	41	0.89 (0.52–1.55)	0.68 (0.26–1.83)	0.455
	>2 hr	45	69	1.88 (1.35–2.61)	1.25 (0.54–2.89)	0.605
Having registered cell phone	Yes	142	495	1	1
	No	87	139	2.13 (1.63–2.78)	1.03 (0.75–1.41)	0.844
Hads a person to contact	Yes	76	397	1	1
	No	153	237	3.01 (2.28–3.96)	1.47 (1.01–2.12)	0.042*
Disclosure status	Disclosed	71	377	1	1
	Not disclosed	158	257	3.12 (2.36–4.14)	1.51 (1.03–2.20)	0.036*
Baseline functional status	Working	101	404	1	1
	Ambulatory	92	182	1.95 (1.47–2.59)	0.93 (0.66–1.33)	0.708
	Bedridden	36	48	3.28 (2.24–4.81)	1.04 (0.63–1.70)	0.877
OI	Yes	154	270	2.59 (1.96–3.42)	1.44 (1.02–2.04)	0.042*
	No	75	364	1	1
Comorbidity other than OI	Yes	47	104	1.68 (1.22–2.31)	0.84 (0.56–1.27)	0.414
	No	182	530	1	1
Baseline WHO clinical stage	Stage I	98	420	1	1
	Stage II	51	104	2.23 (1.59–3.14)	1.45 (0.99–2.12)	0.055
	Stage III	57	81	4.56 (3.27–6.35)	2.09 (1.41–3.12)	0.001**
	Stage IV	23	29	6.00 (3.78–9.50)	2.86 (1.57–5.18)	0.001**
Baseline BMI	<18.5 kg/m2	59	178	0.97 (0.71–1.33)	0.65 (0.45–0.92)	0.017*
	18.5–24.9 kg/m2	116	346	1	1
	≥25 kg/m2	54	110	1.25 (0.90–1.73)	1.23 (0.86–1.76)	0.257
Baseline CD4 count	<200 mm3	58	123	3.13 (2.17–4.52)	1.61 (1.03–2.51)	0.036*
	200–499 mm3	113	207	2.63 (1.92–3.61)	1.54 (1.07–2.22)	0.019*
	≥500 mm3	58	304	1	1
Overall ART initiation time	Within 07 days	176	519	0.76 (0.56–1.03)	1.29 (0.91–1.83)	0.155
	After 07 days	53	115	1	1
IPT status	Received	98	424	1	1
	Not received	131	210	2.63 (2.02–3.41)	1.60 (1.18–2.18)	0.003**
Baseline adherence	Good	92	489	1	1
	Fair	51	59	4.16 (2.95–5.86)	2.80 (1.92–4.09)	0.001**
	Poor	86	86	4.15 (3.09–5.58)	2.51 (1.81–3.47)	0.001**

BMI- body mass index, OI- opportunistic infection, IPT- Isoniazid preventive therapy, 1- reference category, superscript ** indicates p < 0.01, * indicates p < 0.05 and statistically significant categories at multivariable cox analysis at 5% level of significant, CI-confidence interval

Discussion

The incidence density of LTFU observed in this study (9.9 per 100 person-years) was in line with findings from previous studies conducted in, Bahir-Dar City (9.7 per 100 person-years) [31], Gonder (10.9 per 100 person-years) [36] and Kigali city, Rwanda (9.4 per 100 person-years) [37]. However, the rate reported here is higher than that found in several other studies from Ethiopia, including Debre-Markos (3.7 per 100 person-years) [12], Mekele (4.5 per 100 person-years) [38], Arba-Minch (5.3 per 100 person-years) [39] and South Gondar (6.7 per 100 person-years) [17].

Conversely it was lower than that reported in Northwest Gondar (12.26 per 100 person-years) [29], Northwest Amhara (13.45 per 100 person-years) [16], as well as in other African countries such as Malawi (26.0 per 100 person-years) [9] and Liberia (46 per 100 person-years) [40].

The observed incidence suggests that LTFU remains a substantial public health challenge in Eastern Ethiopia despite expansion of ART services under the UTT strategy. The disparities may be explained by differences in healthcare setting, patient tracing systems, follow-up duration, and operational definitions of LTFU. For example, some studies included all facilities providing HIV services, including rural and resource-limited facilities [40], while others were conducted in facilities with poor healthcare access and service quality [16].

Sample size variations from smaller cohorts like the Mekele study (305 participants) [38] to larger cohorts such as the study in Liberia (4185 participants) [40], could also contribute to the observed variation. Additionally, differences in follow-up periods ranging from as short as 2 years [9, 40] to up to 10 years [12], and definitions of LTFU across studies likely contribute to the observed variation. For example, in some studies LTFU was defined as missing ART for more than one month [16], while others defined it as failure to return within 60 days of the last scheduled visit [9].

Differences in health seeking behavior [41], the occurrence of undocumented self-transfer to other facilities without giving prior notice to the caregiving institution and poor patient tracing for early death report which could be considered as LTFU may be additional justificaltion for observed discrepancies [42, 43].

Notably, the relatively lower incidence rate in this study compared to some others might be attributed to the commencing of “Universal test and treat” strategy since 2016 which promotes early treatment initiation. Evidence suggests that early initiation of ART improves retention and significantly reduces the risk of LTFU [44, 45].

The study found the risk of LTFU was significantly higher among HIV positive adults who did not have a registered contact person. Similar findings were reported in Kigali city, Rwanda [37]. The finding highlights the importance of social support and patient tracing mechanisms in HIV care. Individuals without a contact person may have reduced emotional and practical support, making it difficult to maintain regular clinic attendance. In addition, contact persons can assist healthcare providers in tracing patients who miss appointments, particularly in situations where patients lack phone access or become unreachable [46, 47].

Likewise, non-disclosure of HIV status increased risk of LTFU. This finding was corroborated by previous studies conducted in Central, Eastern and Northwest, Ethiopia [13, 31, 48]. A potential reason for this trend could be the fear of stigma and discrimination by intimate partners, family members, or the wider community, that they will not disclose their HIV status and will not get enough social support which could lead them to self-isolation, missed appointments, and ultimately LTFU [49–51]. Moreover keeping status secrete might lead to stress and depression, making challenges to the long-term retainment in the care and later leads to disengagement from the follow-up [52].

Variables such as age, sex, marital status, educational status, and occupation were not significantly associated with LTFU in the present study. This discrepancy may reflect contextual differences in service accessibility and increased ART coverage under the UTT strategy, which may have reduced disparities across socio-demographic groups.

Advanced WHO clinical stage was a strong predictor of LTFU in this study. Participants in WHO clinical stage III and IV had significantly higher hazards of LTFU compared with those in stage I. Similar findings were reported from Debre-Markos, North Shoa, Tanzania, and South Africa [10, 12, 28, 35]. Patients with advanced disease are more vulnerable to opportunistic infections, severe illness, and physical limitations, which may reduce their ability to attend regular clinic visits [53]. In some cases, advanced illness may result in unreported deaths that are later misclassified as LTFU [54].

However, a study conducted at Mizan-Aman General Hospital, Ethiopia, found that Patients with advanced clinical stage III or IV at entry were less likely to be LTFU compared to those in stage I [18]. The discrepancy might be explained by differences in the quality of care provided to advanced stage patients and differences in study populations, in which the Mizan-Aman study includes all age groups, including children and adolescents, who often lives with caregivers, who would play a key role in retention by providing emotional supports and ensuring that appointments are not missed.

Similarly, there was an elevated risk of LTFU among patients that had a baseline CD4 cell count of < 200 cells/mm³ and 200–499 cell/mm³ than those with CD4 cell counts ≥ 500 cells/mm³. This observed association is in parallel with studies conducted in India [22], South Africa [55], and different areas of Ethiopia [18, 28, 34, 48]. This finding could possibly be explained by the fact that patients with a low CD4 cell counts often have compromised immunity, making them more vulnerable to opportunistic infections and more likely to be in advanced clinical stage, which ultimately become too ill to attend regular follow-up visits.

Furthermore, this group of population might have an elevated risk of mortality and due to lack of formal reporting of the death events; these cases may be misclassified as LTFU [56, 57]. However, in despite of most studies agreed on that the risk of LTFU was higher in patients with low CD4 cell count, study done by Mberi et al. showed a lower risk of LTFU among patients with low CD4 cell count [10]. This contrasting finding might be due to the difference in setting, health seeking behavior and service delivery models, including enhanced follow-up and targeted support system for specific patient groups [58].

Contrary to previous studies done in Ethiopia [28, 29, 59], this study found that patients with baseline BMI of <18.5 kg/m² were at a lower risk of LTFU compared to those with a BMI between 18.5 and 24.9 kg/m². The possible explanation for this might be attributed to the additional micronutrient supplementation service given to those undernourished patients by the facilities, which possibly enhances their retention in the care. Furthermore, these patients might have received more intensive counseling and follow-up support contributing to their more retention in the follow-up than of their counterparts. While these factors could partly account for the observed difference, this finding warrants further investigation to clarify the underlying mechanisms and its generalizability across different settings.

The risk of LTFU was higher among patients with OIs compared to those without OIs. This aligns with findings from Gondar, Ethiopia [29], as well as studies from Kembata zone, Central Ethiopia [48] and Togo [60]. The increased risk may be attributed to the challenge of taking many pills due to the OIs, potentially causing pill burden, and in order to overcome this patients might miss doses, abandon their follow-up, and eventual discontinuation of follow-up. Additionally, patients with OIs are at greater risk of adverse drug interaction, toxicity and sever illness, which may further compromise retention in care [61, 62]. In some cases, severe OIs may result in unrecorded deaths, with such cases erroneously identified as LTFU [63].

Moreover, patients who did not receive IPT were at higher risk of LTFU than those who received the therapy. Congruent finding, corroborated by studies conducted in Ethiopia [29, 34, 48], consistently showed that the absence of IPT was associated with an elevated risk of LTFU. Similarly, studies from Mekele and Pawi demonstrated the protective effect of IPT in reducing LTFU [14, 38]. The protective role of IPT may be attributed to its direct benefit in preventing tuberculosis, the most common OI and a major cause of morbidity and mortality among PLHIV. Adherence to national ART guideline and completion of the recommended IPT course may therefore enhance patient retention in care by reducing TB related illness and its associated complications [64, 65].

Regarding ART drug adherence, patients with fair or poor adherence were more likely to become LTFU compared to those with good adherence. Similar observation was demonstrated by different studies conducted in different part of Ethiopia [29, 31, 34] and Tanzania [66]. Another congruous finding was also documented by a study from Arba-Minch with a higher hazard of LTFU among patients with suboptimal adherence [39]. The possible rational behind this association might be explained by the fact that poor adherence can lead to declining CD4 counts, weakening the immune system, and advancing disease stage, potentially rendering patients bedridden and more likely to disengage from care [67]. Furthermore, determinants of poor adherence such as poverty, pill burden, drug side effects, depression, stigma and discrimination may indirectly contribute to LTFU [68, 69].

Although prior studies have demonstrated significant association between LTFU and other factors such as sex [13, 66], age at enrollment [25, 39], marital status [31, 70], educational level [17, 22], occupation [10, 29], access to a registered cell phone [12, 71], baseline functional status [34, 36], comorbidities [28] and ART initiation time after diagnosis [25, 31], these variables were not statistically significant in this study.

Strengths and limitations

A major strength of this study was the use of eight years of follow-up data collected from multiple public hospitals in Eastern Ethiopia, which improves the representativeness and generalizability of the findings. In addition, the study considered important access-related variables such as distance from health facilities and travel time, which are not addressed in previous Ethiopian studies on time to LTFU. The relatively large sample size and application of survival analysis techniques also strengthened the validity of the findings.

However, the study has limitations that need to be put into considerations while interpreting the results. Since the study relied on retrospective secondary data, records with missing ART initiation dates and undefined treatment outcomes were excluded from the analysis, reducing the final number of records included in the study. In addition, some socio-demographic and clinical characteristics may have changed over time, but these time-varying factors were not captured and considered in the analysis. To minimize this limitation, data quality checks was performed during data extraction, and records with missing key outcome variables were excluded from the analysis.

Important factors such as socioeconomic status, substance use, treatment failure and patient satisfaction were not available in patient records and therefore could not be evaluated, which may limit the comprehensiveness and generalizability of the findings. In addition, some patients classified as LTFU may actually have died or self-transferred to other facilities without documentation, potentially leading to overestimation of the true incidence of LTFU.

Conclusions

LTFU was relatively high and remains a substantial challenge among adults receiving ART in Eastern Ethiopia during the UTT era. Clinical severity, poor adherence, and limited social support were important predictors of LTFU. In order to reduce LTFU and improve treatment outcomes, we recommend strengthening patient monitoring, adherence counseling, and targeted support for high-risk patients. In addition further prospective and qualitative studies incorporating socio-economic, patient satisfaction and facility-related factors are needed.

Abbreviations

AHR

Adjusted Hazard Ratio

AIDS

Acquired Immunodeficiency Syndrome

AOR

Adjusted Odd Ratio

ART

Antiretroviral Therapy

ASHR

Adjusted Sub-Hazard Ratio

BMI

Body Mass Index

CD4

Cluster Differentiation 4 Cell

CPT

Cotrimoxazole Preventive Therapy

FMOH

Federal Ministry of Health

HIV

Human Immunodeficiency Virus

HFCSUH

Hiwot Fana Comprehensive Specialized University Hospital

INH

Isoniazid

LTFU

Lost to Follow-up

PLHIV

Patient Living with Human Immunodeficiency Virus

SSA

Sub-Saharan Africa

TB

Tuberculosis

UNAIDS

Joint United Nations Program on HIV/AIDS

UTT

Universal Test and Treat

WHO

World Health Organization

Author’s contributions

MS conceived the idea, analyzed, interpreted the data and drafted the first manuscript. ABW and AT contributed in methodology, analysis and revising of the manuscript. AT, BS, AG, RM, and KY contributed to analysis, and revising the manuscript. Finally, all the authors read and approved the final version of the manuscript.

Funding

This study was funded by Haramaya University. The funding body had no role in the design, data collection, analysis, interpretation, or writing of the manuscript.

Data availability

The datasets used in this study are available from the corresponding author and can be accessible through reasonable request.

Declarations

Ethical

Ethical approval was obtained from the Institutional Health Research Ethics Review Committee (IHRERC) of Haramaya University, College of Health and Medical Science (Reference No. IHRERC/176/2024). Official support letters were submitted to the regional health bureaus and respective hospitals, and written consent was obtained from hospital directors. As the study was based on the retrospective review of medical records, the requirement for individual patient consent was waived. To ensure confidentiality, all data were anonymized, and personal identifiers such as patient names, medical record numbers were not included in the extraction tool. The study was conducted in accordance with the Declaration of Helsinki.

Competing interests

The authors declare no competing interests.