Community outreach programs and major adherence lapses with antiretroviral therapy in rural Kakamega, Kenya

ABSTRACT

We investigated features of major adherence lapses in antiretroviral therapy (ART) at public Emusanda Health Centre in rural Kakamega County, Kenya using medical records from 2008 to 2015 for all 306 eligible patients receiving ART. Data were modelled using survival analysis. Patients were more likely to lapse if they received stavudine (hazard ratio (HR) 2.54, 95% confidence interval (95%CI):1.44–4.47) or zidovudine (HR 1.64, 95%CI:1.02–2.63) relative to tenofovir. Each day a patient slept hungry per month increased risk of major adherence lapse by 3% (95%CI:0–7%). Isolated home visits by community health workers (CHWs) were more effective to assist patients to return to the health centre than isolated phone calls (HR 2.52, 95%CI:1.02–6.20).

KEYWORDS:

• Adherence
• food insecurity
• tenofovir
• support groups
• community health workers
• home visits

1. Introduction

With limited monitoring for resistance of HIV to antiretroviral therapy (ART) in rural Kenya (Brooks et al., 2016), providing adherence support is essential to extend life (Chaiyachati et al., 2014; Grimsrud, Lesosky, Kalombo, Bekker, & Myer, 2016; Hickey et al., 2015; Mills et al., 2014; Yotebieng et al., 2016). We describe features of adherence and programmes to assist with adherence implemented by rural Emusanda Health Centre in Kakamega County, Kenya, between 2008 and 2015.

2. Methods

2.1. Study site and community outreach programmes

Opened in 2008, Emusanda has tested more than 50,000 people for HIV, with more than 2,000 positives, and treats more than 550 patients with HIV free of charge. Monthly support group meetings are led by medical staff and peer educators who train attendees to contact fellow patients who miss appointments. The method of contacting patients at Emusanda is determined by convenience. Nearly half of patients have phone numbers on file, often shared with a relative or neighbour. Eleven community health workers (CHWs) serve the area around Emusanda. Alternately, an HIV-positive support group patient who has a relationship with the missing patient may initiate home visit. After most successful contacts, the missing patient promises to return, and if not returning after a variable period of time, may be contacted again.

2.2. Definition of major adherence lapse

Adherence can be measured by self-report, pill counts, pharmacy records, electronic dose monitoring, drug detection, and viral load (Haberer et al., 2017). Pharmacy records are least invasive and have consistently been shown to be associated with virological failure and mortality (Orrell et al., 2017). At Emusanda and other public dispensaries in Kenya, patients receive medications sufficient only until their next required appointment. If medical staff discover that medications remain from previous visits, medications are reduced accordingly. Patients may request transfer to another dispensary by letter. Without transfer, patients can collect only a few days of medications from other dispensaries. A major adherence lapse was defined as missing a required appointment by at least one month as medical staff agreed that insufficient medications would have been obtained for adherence.

2.3. Data collection

In this chart review at Emusanda between 30 June 2015 and 22 August 2015, data were recorded in Research Electronic Data Capture (Harris et al., 2009) database for all 306 patients who, at the time of the study, were prescribed ART, were at least 18 years, were not expected to be pregnant, and were considered mentally competent by medical staff. All patients provided informed consent during usual visits or after contact by usual means. On day of consent, patients were provided lunch and cost of transportation.

2.4. Ethical approval

Approvals were obtained from Institutional Review Board for Human Subjects at Stanford University; Institutional Ethics Review Committee at Masinde Muliro University of Science and Technology; National Commission for Science, Technology and Innovation, Republic of Kenya; Kakamega County Health Management Committee; Kakamega County Chief Officer for Health Services; County Commissioner for Kakamega County; and Director of Education for Kakamega County.

2.5. Statistical analysis

Using available medical records, twelve variables were included in Cox model for time to first major adherence lapse: gender, age, marital status, number of children, approximate distance of home, WHO clinical stage, history of tuberculosis, nucleoside/nucleotide reverse transcriptase inhibitor (NRTI) medication, non-nucleoside reverse transcriptase inhibitor (NNRTI) medication, average number of days sleeping hungry per month, attendance at support group, and time correction defined as days between patient beginning ART at Emusanda and 23 August 2015 to correct for changes in health centre policy over time. Another Cox model was fit to the data for patients who had been contacted exactly once in the first major adherence lapse prompting contact to find rate ratios for returning to Emusanda following contact. This model included seven variables for 55 patients to ensure sufficient data for number of covariates by sequentially removing covariates that were not statistically significant. Further reduced models with six variables omitting either method of contact or attendance at the support group yielded nearly identical results. All models satisfied model assumptions as confirmed by plots and tests of Schoenfeld residuals. Statistical analyses used Stata 14.1.

3. Results

3.1. Characteristics of patients

Median time between start of ART at Emusanda and end of data collection was 805 days (2.20 years). More than three quarters of patients, 76% (233 of 306), were female. Median age was 39 years (range 18 to 85). 30% (93 of 306) were asymptomatic throughout treatment. Tuberculosis had been diagnosed and confirmed in 12% (36 of 306). 36% (110 of 306) received tenofovir throughout, 33% (100 of 306) received zidovudine throughout, and 31% (96 of 306) switched regimens, mostly (78%, 75 of 96) due to discontinuation of stavudine in 2014. All patients received nevirapine or efavirenz, 70% (213 of 306) nevirapine throughout. Patients lived up to 75 kilometres distant, though 85% (260 of 306) lived within 13 kilometres. Lack of food was widespread with 23% (71 of 306) sleeping hungry at least ten days per month. Only 6% (18 of 306) had never been married, with 42% (130 of 306) in monogamous marriages, 12% (37 of 306) in polygamous marriages, 24% (72 of 306) widowed, and 16% (49 of 306) separated. Only 4% (11 of 306) had no children and 34% (105 of 306) had 5 or more. Data were not requested for two patients not considered mentally competent by medical staff, four who were pregnant, and 35 who were younger than 18 years.

3.2. Characteristics of major adherence lapses

Of patients, 45% (137 of 306) had a major adherence lapse with median of 293 days after beginning ART. From the Kaplan-Meier curve in Figure 1(a), 25% were expected to have a major adherence lapse within 296 days (0.81 years) and 50% within 885 days (2.42 years). In Figure 1(b), gender did not show significant difference by log-rank test for equality of survivor functions (p = 0.2330). Median duration of first major adherence lapse was 54 days (range 28 to 518).

Figure 1. Kaplan-Meier survival curves for time without major adherence lapse. (a) Kaplan-Meier survival without major adherence lapse. (b) Kaplan-Meier survival without major adherence lapse by gender.

3.3. Risk factors for major adherence lapses

Patients were more likely to lapse receiving stavudine (hazard ratio (HR) 2.54, 95% confidence interval (95%CI):1.44–4.47) or zidovudine (HR 1.64, 95%CI:1.02–2.63) relative to tenofovir (Table 1). As days sleeping hungry increased, lapses were more likely (HR 1.03, 95%CI:1.00–1.07 per day). Patients in monogamous marriages were more likely to lapse than patients in polygamous marriages (HR 2.27, 95%CI:1.17–4.40). Patients in WHO clinical stage IV were less likely to lapse than those in stage II (HR 0.26, 95%CI:0.08–0.85). Patients with confirmed tuberculosis were more likely to lapse (HR 1.99, 95%CI:1.03–3.84).

Table 1. Hazard ratios for major adherence lapses.

	Hazard ratio (95% CI)	p-value
Gender
Female (reference)	1.00	–
Male	0.63 (0.39–1.02)	0.0601
Age	0.99 (0.97–1.00)	0.1416
WHO clinical stage
II (reference)	1.00	–
I	0.94 (0.62–1.42)	0.7574
III	0.87 (0.51–1.49)	0.6088
IV	0.26 (0.08–0.85)	0.0261*
Confirmed tuberculosis	1.99 (1.03–3.84)	0.0407*
NRTI
Tenofovir (reference)	1.00	–
Zidovudine	1.64 (1.02–2.63)	0.0420*
Stavudine	2.54 (1.44–4.47)	0.0012*
NNRTI
Nevirapine (reference)	1.00	–
Efavirenz	1.00 (0.57–1.74)	0.9929
Distance of home in kilometres	1.00 (0.98–1.01)	0.7206
Days sleep hungry per month	1.03 (1.00–1.07)	0.0423*
Marital status
Married polygamous (reference)	1.00	–
Widowed	1.42 (0.70–2.88)	0.3347
Separated	1.77 (0.85–3.65)	0.1246
Never married	1.86 (0.65–5.38)	0.2499
Married monogamous	2.27 (1.17–4.40)	0.0155*
Number of children	0.99 (0.89–1.10)	0.8614
In support group	1.11 (0.70–1.74)	0.6572
Time correction	1.00 (1.00–1.00)	0.0148*

*p-value is less than 0.05.

3.4. Return from major adherence lapses

Of 55 patients successfully contacted exactly once followed by return, 25 were contacted by phone call by medical staff, 18 were contacted by home visit by CHW, and 12 were contacted by home visit by fellow patient. Median time to return following phone contact was 18 days (range 1 to 179), CHW contact 3 days (range 1 to 50), and fellow patient contact 10.5 days (range 1 to 36). Home visits by CHW led to improved return (HR 2.52, 95%CI:1.02–6.20) relative to successful phone call (Table 2). Patients attending the support group within the previous year were more likely to return when contacted (HR 2.16, 95%CI:1.02–4.54).

Table 2. Hazard ratios for return to the health centre with seven variables.

	Hazard ratio (95% CI)	p-value
Mode of contact
Telephone (reference)	1.00	–
Home visit by CHW	2.52 (1.02–6.20)	0.0442*
Home visit by fellow patient	1.73 (0.65–4.60)	0.2703
WHO clinical stage
II (reference)	1.00	–
I	1.21 (0.49–3.04)	0.6773
III	1.33 (0.59–2.99)	0.4909
IV	5.17 (0.49–54.86)	0.1730
Distance of home in kilometres	1.00 (0.98–1.03)	0.7444
Days sleep hungry per month	0.99 (0.92–1.05)	0.6670
Marital status
Married monogamous (reference)	1.00	–
Never married	1.29 (0.31–5.37)	0.7233
Separated	0.76 (0.27–2.18)	0.6150
Widowed	0.49 (0.22–1.09)	0.0811
Married polygamous	0.53 (0.19–1.46)	0.2158
In support group previous year	2.16 (1.02–4.54)	0.0436*
Time correction	1.00 (1.00–1.00)	0.0254*

*p-value is less than 0.05.

4. Discussion

In this study in rural Kenya, half of patients were expected to have a major adherence lapse within two and a half years of beginning ART. Tenofovir, food security, and advanced illness were associated with extended time before first major adherence lapse.

Prevalence of HIV among women in Kenya is higher than among men (6.3% for women 15–49 years of age in 2015; 5.5% for men 15–49 years of age in 2015 [National AIDS Control Council, 2016]) likely due to greater physiologic vulnerability to HIV infection (Ramjee & Daniels, 2013). However, the much larger percentage of female patients on ART may be explained by programmes that focus on women (Camlin et al. 2016) often during pregnancy (Gunn et al., 2016). Other possible reasons for the large gender difference are increased loss to follow-up among men (Ochieng-Ooko et al., 2010) and higher mortality among men (Mills et al., 2011) even with similar adherence (Boullé et al., 2015; Hawkins et al., 2011). Medical staff at Emusanda have noted high stigma among men regarding testing and treating of HIV as also observed in Zimbabwe (Skovdal et al., 2011).

Marital statuses recorded by the medical staff follow the required template from Ministry of Health and reflect social structures prevalent in rural agrarian communities of Kenya (Owili et al., 2016). Nearly one in four patients are widowed, reflecting the high mortality rate, and is a major cause of elevated poverty (Ismail, Eisa, & Ibrahim, 2016). Nearly one in six patients are separated which typically involves the wife returning to her parents’ ancestral land with relative loss of possessions. Nearly one in eight patients are in polygamous marriages where the senior wife maintains her home and children on the husband’s ancestral land, at times following marital differences, while the husband builds a separate house to live with his junior wife. These social structures do not determine number of sexual partners for either men or women (Kioko, 2015).

Nutritional supplementation is important for patient adherence (Weiser et al., 2010). Practical ideas for nutritional supplementation in Kenya have successfully been implemented (Ernst, Ettyang, & Neumann, 2014; Hong et al., 2012; Lim et al., 2009; Mamlin et al., 2009; Nagata et al., 2014), though government and donor funding commitments are needed to extend nutritional programmes more broadly (Haberer et al., 2017).

Acknowledgements

We thank CHWs Mildred Alibitsa, Sophia Amakove, Getreil Andayi, Odelia Arumba, Dorcas Bulima, Winfred Chengo, Josphine Eruata, Leonida Imbosa, Chrisandos Odinga, and Beatrice Shikami; Elders Chair Josphat Sakwa; Emusanda Chair Stanley Ingoka; Dr. James Ouda; Prof. Elizabeth Abenga; Prof. Eran Bendavid; and Prof. Lars Osterberg for their contributions.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This research was supported by Stanford Medical Scholars Fellowship [grant number 11571].