Outcomes and associated risk factors of patients traced after being lost to follow-up from antiretroviral treatment in Lilongwe, Malawi

The Lighthouse Trust (Lighthouse) is part of Kamuzu Central Hospital, the tertiary health facility of Malawi's central region within the capital Lilongwe and has offered HIV counseling and testing, clinical and home-based care since 2002 [14]. From 2002 to 2004, Lighthouse was the only public-sector ART clinic for the central and northern region and about 60 patients started each month on a fee-basis ("paying period"). In June 2004, the Ministry of Health opened 60 ART clinics in the public sector throughout Malawi, and Lighthouse began offering free ART, with a target of starting 150 new patients on ART per month.

Lighthouse has maintained electronic data for all initial and follow-up visits since its inception. Details of procedures for clinic registration and ART initiation during the "paying period" have been described elsewhere [15]. In line with the 2003 national ART-guidelines, Lighthouse implemented standardized procedures for patient management and monitoring[10]. In short, HIV-infected patients register at the clinic with name, age, sex, place of residence, phone number and are asked consent to be contacted if they miss their appointment. After registration, clinicians treat active opportunistic diseases and provide cotrimoxazole prophylaxis. Patients in WHO stage 3 and 4 or with a CD4 count of below 200 cells/μl (until 2005) or below 250 cells/μl (from 2006) are eligible for ART. For children (as defined as persons 0-14 years old at the time of ART initiation) the age-dependent absolute CD4 count value thresholds for severe immunosuppression were used. The majority of children started at an age, when adult normal values are reached. To initiate ART, patients are required to attend a group education session and then are started on a fixed dose combination regimen of d4T/3TC/NVP with a two week lead-in phase. Thereafter, patients are seen monthly and receive an ART supply for 30 days. Pill counts are recorded at each visit to assess adherence. Appointments are given according the remaining pills and the new supply. After attending a refresher adherence counseling session, patients are eligible to receive a supply for 60 days and visit the clinic every 2 months. If acutely ill, patients are referred to Kamuzu Central Hospital wards. Clinic staff updates outcomes on the patient master card at every visit and the ART register is updated at least quarterly.

Between August and December 2005 we actively traced patients LTFU, i.e. patients who were at least two weeks late for a scheduled appointment according to their last ART supply. We thus included patients with shorter delays than 60 days proposed by the Malawian national program guidelines as to evaluate the effect of earlier tracing on patient outcomes. The study population was restricted to all patients ever starting ART at the Lighthouse clinic until 31^st of May 2005. We verified contact information (phone and place of residence) and whether patients consented for external tracing; Lighthouse began collecting this contact information in May 2004, so patients already LTFU by this time were not traced. We excluded patients living outside Lilongwe city without a phone number, patients living within Lilongwe city but without address and phone number and those who did not consent for tracing.

Using the list of eligible patients, Kamuzu Central Hospital death registry was cross-checked for Lighthouse patients and their date of death, and the ART register at Mzuzu Central Hospital in the northern region of Malawi, which opened its ART clinic shortly before start of the study, was searched for Lighthouse patients LTFU whose transfer out was not registered.

Study assistants contacted the patients, family members or contact persons by phone or in-person. A brief survey instrument was used to collect vital status and relevant information on current ART use, current site of care, or date of death. All study data were collected on paper forms and subsequently entered into an electronic database.

We used descriptive analyses to examine demographic and clinical characteristics and tracing history of patients LTFU and their outcomes. Outcomes were classified as patient alive, died or untraceable. Patients found alive were further categorized as on ART, transferred out to another ART program or having stopped ART.

We performed logistic regression for both successful tracing and mortality among successfully traced patients to determine risk factors for these outcomes and, in a second step, performed a sensitivity analysis restricted to patients living in Lilongwe. Predictor variables for both models included gender, age at start of ART (in years; groups 15-29; 30-39; 40-49; ≥ 50), time period of starting of ART (2 groups: until June 2004, paying for ART; June 2004 or later, free ART), reason for starting ART (4 groups: low CD4 count; WHO stage 3; WHO stage 4; other or unknown), follow-up time from start of ART until the last patient contact (time of last visit in months: no follow-up; 1-6; 7-12; 13 or later) and providing a telephone contact. We evaluated adjusted odds ratios that were mutually adjusted for all other predictor variables.

The day and month of death was commonly unknown for patients who were traced outside of the health facilities. To evaluate mortality rates since ART initiation in patients LTFU with known outcomes (i.e., those traced successfully) we performed sensitivity analysis by stepwise increasing the presumed date of death by 3 months between one day after the last contact date and the date of successful tracing for that patient. Under these different scenarios we evaluated mortality among successfully traced patients using Kaplan-Meier analysis over months 1 to 12 and months 13 to 24. Rates are expressed in deaths per 100 person-years. All analyses were done using Stata (version 10.1; Stata Corporation, College Station, Texas, USA).

The study has ethical approval from the National Health Science Research Committee (NHSRC), the national ethics committee in Malawi. General measures were provided in all ART facilities to ensure patient confidentiality, consent for HIV testing and counseling, and support for those who received a positive HIV test result. Data collected for this study did not include personal identifiers. Since the study was partly funded by the U.S. Centers for Disease Control and Prevention (CDC) and involved a co-investigator from CDC, the protocol had to undergo ethics review by CDC. As the study design and purpose aims at evaluating and improving services, it was approved and classified as program evaluation, a type of public health non-research according to CDC regulations [16].

Until 31^st May 2005, 3846 patients had ever started ART at the Lighthouse and 1840 (48%) were LTFU, of whom 1800 (98%) had given consent to tracing (Figure 1). For 1076 patients the contact address was outside Lilongwe and no phone contact was given or was within Lilongwe and neither phone number nor address was given. From the list of the patients remaining eligible for tracing, a review of records in Lilongwe and Mzuzu Central Hospitals revealed vital status of 93 patients, so that 631 patients had to be traced by outreach, either by phone or in-person. In total we thus evaluated outcomes of 724 patients LTFU that included 659 adults and 65 children less than 15 years of age, representing 40% (724/1800) of all patients LTFU who gave consent for tracing.

Table 1 describes the baseline demographic and clinical characteristics of the study population. Adults and children had a median age of 36 and 6 years, and the percentage of females was 53% and 43%, respectively. The median year of ART start was 2004 and start of ART was free of charge for most adults (67%) and children (80%). The median (IQR) baseline CD4 counts were 71 cells/μl (22-141) and 184 cells/μl (29-423) and the most common reason for ART start was WHO stage 3 in 67% and 89% of adults and children, respectively. The median (IQR) duration of follow-up in the ART-clinic was 73 days (14-303) and 92 days (13-210), and 67% and 69% of adults and children were LTFU within 6 months after ART start. Telephone contact was available for 47% of adults and 37% of children and 86% of adults and 89% of children were living in Lilongwe.

Of the 631 patients traced by outreach, 55 (9%) patients had two and 2 patients had three successive tracing attempts. Altogether, 389 (57%) tracing attempts were performed in-person and 299 (43%) by phone. First tracing attempts were successful in 61% of the in-person attempts (204/336) and 73% (214/295) by phone attempts. Tracing success declined from 66% (418/631) to 36% (20/55) from first to second tracing attempt (data not shown).

The updated outcomes after tracing are given in Table 2. In total, of those eligible for tracing, 534 (74%) patients LTFU were successfully traced. These included 486 adults and 48 children, of whom 201 (41%) and 16 (33%) respectively, were found to be dead. Among the 317 patients found to be alive, 285 (90%) were still on ART (157 at Lighthouse and 128 at another ART clinic) and 32 (10%) had stopped ART.

The percentage successfully traced was 81.4% and 67.0% (P < 0.001) for patients with and without a phone contact, respectively. Controlling for gender, age, treatment period, reason for start of ART and time of the last clinic visit from ART start, having a phone number doubled the odds of successful tracing (aOR 2.07, 95% CI: 1.42-3.01, P < 0.001) (Table 3). Other factors, including reasons for start of ART (P = 0.39) and follow-up time until the last clinic visit (P = 0.94) were not associated with tracing success. After restricting the analysis to patients living in Lilongwe the results stayed mostly the same and availability of a phone number remained the major risk factor (data not shown).

Mortality rates in successfully traced patients between months 1 and 12 after start of ART were 58 (95% CI: 50-67), 50 (95% CI: 43-58) and 45 (95% CI: 38-53) per 100 person-years, when allowing a maximal time between the date of last visit and the presumed death date of 3, 6 and 9 months, respectively. Between months 13 and 24 the respective mortality rates were 19 (95% CI: 12-30), 20 (95% CI: 13-31) and 27 (95% CI: 19-39) per 100 person-years.

Table 4 shows the crude percentage mortality and adjusted odds ratios of death in relation to risk factors. The percentage of deaths was 64.6% in patients with no follow-up following ART initiation as compared to 50%, 16.7% and 19.8% in those with a last visit in months 1 to 6, months 7 to 12 and months 13 or later, respectively. The corresponding adjusted odds ratios were 0.55 (95% CI: 0.31-0.96), 0.073 (95% CI: 0.03- 0.18), and 0.058 (95% CI: 0.024-0.14), indicating declining mortality with increasing follow-up time (P < 0.0001). The proportion of recorded deaths was higher among successfully traced patients with a phone contact (43.6%) as compared to those without a phone (39%; aOR 1.79, 95% CI: 1.19-2.71, P = 0.005). The adjusted odds of death were further higher among patients that initiated ART during the paying period as compared to the free period [aOR = 2.28 (1.10-4.72); P = 0.022]. Gender, age group and reason for start of ART were not strongly associated to the risk of death in successfully traced patients. The sensitivity analysis restricting to patients living in Lilongwe did not reveal major changes of the results and time of last visit since ART start and availability of a phone number remained the major risk factors for mortality among successfully traced adult patients (data not shown).