Objective To compare the effects of low-level facility-based interventions on patient retention rates for cardiovascular (CV) disease in an environment of task shifting and nurse-led care in rural health districts in Central Cameroon.
Methods This study is an open-label, three-arm, cluster-randomised trial in nurse-led facilities. All three groups implemented a treatment contract. The control group (group 1) had no additional intervention, group 2 received the incentive of 1 month of free treatment every forth month of regularly respected visits, and group 3 received reminder letters in case of a missed follow-up visit. The primary outcome was patient retention at 1 year. Secondary outcomes were adherence to follow-up visit schemes and changes in blood pressure (BP) and blood glucose levels. Patients’ monthly spending for drugs and transport was calculated retrospectively.
Results A total of 33 centres and 221 patients were included. After 1 year, 109 patients (49.3%) remained in the programme. Retention rates in groups 2 and 3 were 60% and 65%, respectively, against 29% in the control group. The differences between the intervention groups and the control group were significant (P < 0.001), but differences between the two intervention groups were not (P = 0.719). There were no significant differences in BP or fasting plasma glucose trends between retained patients in the study groups. Average monthly cost to patients for antihypertensive medication was € 1.1 ± 0.9 and for diabetics €1.2 ± 1.1. Transport costs to the centres were on average €1.1 ± 1.0 for hypertensive patients and €1.1 ± 1.6 for patients with diabetes.
Conclusions Low-cost interventions suited to an environment of task shifting and nurse-led care and needing minimal additional resources can significantly improve retention rates in CV disease management in rural Africa. The combination of a treatment contract and reminder letters in case of missed appointments was an effective measure to retain patients in care.
Objectif: Comparer les effets des interventions de base dans les établissements sur les taux de rétention des patients souffrant de maladies cardiovasculaires dans un contexte de délégation des tâches et des soins, dirigé par des infirmiers dans les districts de santé ruraux dans le centre du Cameroun.
Méthodes: Essai randomisé en grappes, ouvert, à trois bras, dans les établissements dirigés par des infirmiers. Les trois groupes ont tous établi un contrat de traitement. Le groupe témoin (groupe 1) n’a pas reçu d’intervention additionnelle, le groupe 2 a reçu la mesure incitative consistant en 1 mois de traitement gratuit après 4 mois de visites régulièrement respectées et le groupe 3 a reçu des lettres de rappel en cas de manque de la visite de suivi. Le critère principal était la rétention des patients à 1 an. Les critères secondaires étaient l’adhésion aux schémas des visites et les changements dans la pression sanguine (PS) et les taux de glucose dans le sang. Les dépenses mensuelles des patients pour les médicaments et le transport ont été calculées de manière rétrospective.
Résultats: 33 centres et 221 patients ont été inclus. Après 1 an, 109 patients (49,3%) restaient dans le programme. Les taux de rétention dans les groupes 2 et 3 étaient de 60% et 65% respectivement, contre 29% dans le groupe témoin. Les différences entre les groupes d’intervention et le groupe témoin étaient significatives (p <0,001), mais les différences entre les deux groupes d’intervention ne l’étaient pas (p = 0,719). Il n’y avait pas de différences significatives dans la PS ou dans les tendances du glucose plasmatique à jeun entre les patients retenus dans les groupes d’étude. Le coût mensuel moyen aux patients était de 1,1 ± 0,9 € pour les médicaments antihypertenseurs et de 1,2 ± 1,1€ pour ceux du diabète. Les frais de transport vers les centres étaient en moyenne de 1,1 ± 1,0 € pour l’hypertension et de 1,1 ± 1,6€ pour les patients diabétiques.
Conclusions: Des interventions à faible coût adaptées à un environnement de délégation des tâches et dirigé par des infirmiers de soins nécessitant un minimum de ressources supplémentaires peuvent améliorer considérablement les taux de rétention dans la prise en charge des maladies cardiovasculaires en Afrique rurale. La combinaison d’un contrat de traitement et des lettres de rappel en cas de rendez-vous manqués a été une mesure efficace pour retenir les patients pris en charge.
Objetivo: Comparar los efectos de las intervenciones en emplazamientos de bajo nivel sobre las tasas de retención de pacientes con enfermedades cardiovasculares, en un ambiente con rotación de tareas y liderado por enfermeras, en distritos sanitarios rurales del centro de Camerún.
Métodos: Ensayo abierto, de tres brazos, en conglomerados y aleatorizado en centros liderados por enfermeras. Los tres grupos implementaron un contrato de tratamiento. El grupo control (Grupo 1) no recibió una intervención adicional, el Grupo 2 recibió el incentivo de 1 mes de tratamiento gratis cada cuarto mes de visitas regulares respetadas y el Grupo 3 recibió cartas recordándoles en caso de haberse saltado alguna visitas de seguimiento programada. El resultado principal fue la retención del paciente tras 1 año. Resultados secundarios eran la adherencia a los esquemas de visitas de seguimiento y cambios en la presión (PS) y niveles de glucosa en sangre. Los gastos mensuales de los pacientes en medicación y transporte se calcularon de forma retrospectiva.
Resultados: Se incluyeron 33 centros y 221 pacientes. Tras 1 año, 109 pacientes (49.3%) continuaban en el programa. Las tasas de retención en los grupos 2 y 3 eran 60% y 65% respectivamente, frente a 29% en el grupo control. Las diferencias entre los grupos de intervención y el grupo control eran significativos (p<0.001) pero las diferencias entre los dos grupos de intervención no lo eran (p=0.719). No había diferencias significativas en las tendencias de la PS o en la glucosa en ayunas entre los pacientes retenidos en los diferentes grupos de estudio. El coste medio mensual para los pacientes en medicación antihipertensiva era de € 1.1 ± 0.9 y para los diabéticos €1.2 ±1.1. Los costes de transporte a los centros eran en promedio €1.1 ±1.0 para los pacientes hipertensos y €1.1 ±1.6 para los diabéticos
Conclusiones: Las intervenciones de bajo coste aplicadas en un ambiente con rotación de tareas y liderado por enfermeras, que requieren recursos adicionales mínimos, puede mejorar significativamente las tasas de retención en el manejo de la enfermedad cardiovascular en África rural. La combinación de un contrato de tratamiento y cartas recordando las visitas en caso de que se haya fallado alguna fueron medidas efectivas para retener a los pacientes.
The burden of non-communicable chronic diseases (NCCD) such as hypertension and diabetes is growing rapidly in low- and middle-income countries (Alwan & MacLean 2009). In Sub-Saharan Africa, NCCDs are expected to overtake infectious diseases to become the major cause of death during this decade (Gaziano et al. 2006; WHO 2007; Miranda et al. 2008). Already, 80% of deaths worldwide from cardiovascular (CV) diseases occur in low- and middle-income countries (WHO 2008).
Many of the challenges in tackling CV diseases are similar in developed and emerging economies. As treatment, including lifestyle changes, is for life and the diseases tend to be asymptomatic, dropout rates from treatment programmes are generally very high and adherence to medication correspondingly low (Bovet et al. 2008; Katz et al. 2009; Kengne et al. 2009a,b).
There are few studies on CV diseases in Africa, and most adherence studies in sub-Saharan Africa have been conducted in urban or mixed settings. We are not aware of a trial designed to improve retention of patients with NCCD in rural Africa. Data from observational studies on CV diseases suggest that adherence and retention may be improved by better provider–patient communication, education about the disease and reduction in financial barriers (Maro & Lwakatare 1997; Ohene Buabeng et al. 2004; Akpa et al. 2005; Amira & Okubadejo 2007; Magadza et al. 2009).
In a primary healthcare programme in eight Cameroon districts, where the task to care for hypertensive patients and patients with diabetes has been shifted to nurse-led facilities, our monitoring data indicate a 1-year retention of 18.1% (Labhardt et al. 2010). We here report the results of a randomised trial in this environment of nurse-led care, comparing two low-level facility-based interventions: reminders and financial incentives. The primary outcome was patient retention at 1 year. Secondary outcomes were adherence to follow-up visit schemes and changes in blood pressure (BP) and blood glucose levels. Treatment and transport cost to patients were also assessed.
Study design, location and population
The study was an open-label, three-arm, cluster-randomised trial with one health centre representing one cluster. The 33 centres were allocated to three equal groups. All three groups implemented a treatment contract where patients and providers commit themselves to long-term care for hypertension and/or diabetes. The control group (group 1) had no additional intervention. The incentive group (group 2) received in addition the financial incentive of 1 month of free treatment for patients who regularly attended follow-up visits. The letter group (group 3) received reminder letters in case of a missed follow-up visit.
The study was conducted in five predominantly rural health districts in Central Cameroon. Since 2007, a programme aiming at integrating care for chronic non-communicable diseases into the routine of nurse-led primary health care has been running in the study area. From 2007 to 2008, functional health centres were equipped and trained in care for hypertension and type 2 diabetes. For this study, exclusively nurse-led health centres were eligible to participate if staffed, equipped and trained to care for hypertension and type 2 diabetes at the time of study initiation. Adult patients were eligible for participation if they were newly diagnosed with uncomplicated hypertension and/or type 2 diabetes during the recruitment phase (August 2008 to January 2009) and resident in the catchment area of the respective health centre.
Ethics approval was obtained from the national ethics committee of Cameroon (authorisation No 059/CNE/DNM/08). Patients participating gave written and oral consent and could exit the study at any time while continuing to receive the same care at the same facility. Data were processed confidentially. The trial is registered on http://www.clinicaltrials.gov (NCT00744458).
Allocation, randomisation and interventions
The eligibility of health centres was assessed using data from routine supervision reports. Each clinic in the five districts that was equipped and trained to care for hypertension and diabetes was eligible. The head nurses of the 33 eligible (of 47 registered) nurse-led facilities in the five districts were invited to take part in the study at a meeting with the investigators, where the study purpose and design were presented. Upon accepting the invitation, health centres were allocated to one of the three study groups using sealed envelopes chosen at random by each head nurse.
Subsequently, all nurses in all groups participated in a 4-h training session on how to establish a treatment contract with hypertensive patients or patients with diabetes. Such contracts constituted the background intervention in all study groups. Contracts were set up by the nurses for each newly detected patient with hypertension or diabetes enrolled into the trial. The process includes four sequential steps: (i) an in-depth counselling on the chronic nature of the disease, the importance of a long-term care and its cost; (ii) assessment of the patient’s motivation for long-term care in the current facility; (iii) the identification of possible obstacles in the patient’s psychosocial condition that might interfere with a long-term care; and (iv) a formal commitment signed by patient and nurse, in which patients commit to do their best to adhere to follow-up schemes and nurses commit to do their best to provide good care.
Group 1 served as a control group, with no additional interventions. In group 2, in addition to the treatment contract, patients were offered the financial incentive of 1 month’s free treatment (drugs and blood glucose measurements) for every 4 months of consecutively attended follow-up visits. In group 3, in addition to the contracts, nurses could send a standardised letter inviting patients to consultation in case of a missed visit. A member of the local health committees (comité de santé, COSA) brought this letter in a closed envelope to the house of the patient. Only delivery of the letter was required, and no in-person meetings with the patient or tracing of missing patients was requested.
The primary outcome was retention in the programme, defined as completing at least 12 follow-up visits within 12 months and with the last recorded consultation at least 1 year after the start of the treatment. The first secondary outcome was a composite endpoint of remaining in the study and attending ≥80% of visits within ±7 days of scheduled visit dates. Second secondary endpoints were changes in BP and/or blood glucose levels and control rates among retained patients. Changes were assessed as the difference between the consultation when the treatment was started and the consultations at 6 and 12 months.
Monthly costs were estimated descriptively, based on 1 Euro (€) = 650 CFA. Costs to patients in rural Cameroon are as follows: five CFA per tablet of 50 mg hydrochlorothiazide (HCTZ), 10 CFA per tablet of 10 mg nifedipine, 15 CFA per tablet of 500 mg metformin and 10 CFA per 5 mg glibenclamide. Costs of therapy were estimated based on the data from the last recorded visit for those patients retained at 1 year.
Study procedure and data processing
Head nurses were advised to assess all newly detected patients with hypertension and/or diabetes for participation during a recruitment period from August 2008 to January 2009. As regular communications with some of the very remote facilities are not possible, the recruitment period was pre-specified to 6 months, assuming that every clinic could enrol at least 10 patients during this time. Nurses were supplied with patient-record-sheets that were collected for data entry in March 2009 and February 2010. The predefined end of the study was 31 January 2010. All data were double-entered into EpiData 3.1® (available at: http://www.epidata.dk), and any entry mismatch was corrected manually from the original patient records.
Sample size was calculated for the primary outcome. Based on 1-year retention rates of 18.1% found in the analysis of routine monitoring data from these facilities, we decided that a relevant intervention should lead to a 1-year retention of ≥50%. Applying the formula of Hayes and Bennett for cluster-randomised trials (Hayes & Bennett 1999) using a significance level of 5%, a statistical power of 80% and an estimated between-clinic coefficient of variation of 0.35, a sample size of nine clinics per arm with 10 patients per clinic was calculated.
To test for differences in patient characteristics between the study groups, one-way anova was used. For the primary outcome, patients were analysed as allocated. Crude primary and crude first secondary outcomes were analysed with the chi-square test. Crude and cluster-effect-adjusted hazard ratios (HR) for the primary outcome were calculated using a Cox proportional hazard model. For the changes in BP and blood glucose, two-sided t-test was used for the longitudinal difference and one-way anova for differences in changes between the groups. Logistic regression analysis was used to assess for possible associations between patient characteristics and the primary outcome. All reported P-values are based on two-sided tests and confidence intervals on a coverage probability of 95%. Comparisons with earlier retention data from the same region (2007–2009 cohort) were purely explorative. All analyses were run on stata 10.1®.
Role of the funding source
The programme of nurse-led care for hypertension and diabetes in the study area receives financial support by the two Swiss cantons of Jura and Geneva as well as by the private Swiss foundation Coopération Afrique. The study was sponsored by the KFPE (Commission for Research Partnerships with Developing Countries, http://www.kfpe.ch). Sponsors of the study had no role in study design, data collection, data analysis, data interpretation or writing of the report. Staff at the participating facilities as well as the three local trial investigators and co-authors (JRB, MN and EM) received allowances for additional workload caused by the trial.
Figure 1 shows the flow of randomised facilities and recruited patients. Three facilities of the incentive group were excluded from the study during the recruitment period. All trained staff at two centres were reallocated to other facilities outside the study area during the recruitment period, and one centre recruited no patients. Table 1 displays the characteristics of the remaining 30 facilities and their head nurses. Most facilities were public and had a rather low utilisation rate per year. Nurses of the control group were slightly older and had a better performance in the knowledge assessment test as compared to the intervention groups (Table 1). During the recruitment period, 229 patients with newly detected diabetes and/or hypertension were assessed for inclusion in the study. Six patients refused participation. Two patients erroneously diagnosed with diabetes were excluded retrospectively from the analysis. The analysis population included all 221 remaining patients (median 7 per facility, interquartile range 5–9). Tables 2 and 3 display the sociodemographic and clinical characteristics of the study population. Clinical and socio-demographic characteristics of the study groups were similar.
Table 1. Characteristics of the 30 NPC-facilities and their head-nurses participating in the study. Performance in the MC test was assessed with a multiple-choice questionnaire on diabetes and hypertension that the nurses had to fill in at the end of the study period. Continuous variables are displayed as mean-values ± SD. P-values derive from one-way anovas
Control (group 1) (11)
Incentive (Group 2) (8)
Letter (Group 3) (11)
Type of institution
Utilisation (visits per year)
1291 ± 1291
813 ± 1028
1263 ± 1173
Sex of the head-nurse
Age of the head-nurse
49 ± 6
42 ± 7
46 ± 4
Medical education of head-nurse
Registered nurse (3 years)
Staff nurse (2 years)
Nurse assistant (1 year)
Professional experience of head-nurse
23 ± 10
16 ± 7
21 ± 5
Performance of head-nurse in MC knowledge test
93% ± 9
88% ± 12
75% ± 17
Table 2. Socio-demographic indicators of patients participating in the study. For continuous variables the mean ± SD is indicated. P-values are derived from one-way anovas
Control (group 1) (n = 89)
Incentive (Group 2) (n = 55)
Letter (Group 3) (n = 77)
59.9 ± 11.8
58.7 ± 14.1
60.8 ± 12.2
Civil status (%)
Age at marriage
31.1 ± 12.4
26.9 ± 12.0
31.9 ± 11.7
No. of children alive
5.9 ± 3.6
4.8 ± 4.1
6.3 ± 3.8
Educational level (%)
Type of house (%)
Electricity in the house (%)
Water in the house or on the yard (%)
Possesses a mobile phone (%)
TV in the house (%)
Regular household income (%)
Household income (€)
45 ± 67
91 ± 168
46 ± 60
Distance to facility (km)
6.8 ± 11.0
6.7 ± 9.5
5.1 ± 6.2
Transport cost to facility (€)
1.2 ± 1.3
1.1 ± 1.3
0.8 ± 0.6
Table 3. Clinical baseline-characteristics of patients
Control (Group 1) (n = 89) (%)
Incentive (Group 2) (n = 55) (%)
Letter (Group 3) (n = 77) (%)
Arterial hypertension alone
Arterial hypertension and IFG
Overweight (BMI 25–29.9 kg/m2)
Obesity (BMI >30 kg/m2)
Abdominal obesity (f > 88 cm; m > 102 cm)
At 1 year, 109 (49.3%) of the 221 patients remained in the programme. Figure 2 shows retention rates in the three study groups. For comparison, the figure also shows observational data from a 2007–2009 cohort followed in the same area with routine monitoring only. At 1 year, the retention rates in the incentive and letter group were 60% and 65%, respectively, compared with 29% in the control group. The differences between the intervention groups and the control group were significant (95% CI: 21% to 46%; P < 0.001), but differences between the two intervention groups were not (95% CI: −12% to 22%, P = 0.719). The HR of loss to follow-up was significantly lower in the intervention groups than in the control group: 0.55 in the incentive group (95% CI 0.32–0.97, P < 0.05 vs. control) and 0.36 in the letter group (95% CI 0.23–0.57, P < 0.001 vs. control). The differences remained significant after adjustment for cluster effect, (incentive vs. control: HR 0.44, 0.27–0.72; P < 0.001; letter vs. control: HR 0.38, 0.24–0.61, P < 0.001). In logistic regression analysis of all 221 patients, none of the patient indicators listed in Tables 2 and 3 were significantly associated with remaining in the programme at 1 year.
Overall, 57 (25.8%) subjects remained in the programme at 1 year with timely attendance at ≥80% of scheduled visits, 10%, 38% and 35% in control, incentive and letter group, respectively (95% CI incentive vs.control: 14–42%, P < 0.001; letter vs. control: 13–37%, P < 0.001; incentive vs.letter: −14% to 20%, P = 0.357).
There were no significant differences in BP or fasting plasma glucose changes between retained patients in the three study groups. Among the 104 hypertensive patients retained at 1 year, 72 (69%) had a BP of ≤140/90 mmHg at the last visit. Overall average systolic BP decreased from 175.8 to 135.6 mmHg (95% CI: 35.0–45.4, P < 0.001) and diastolic from 100.7 to 80.1 (17.3–23.9; P < 0.001).
Few patients with diabetes remained in the study at 1 year (n = 15). Among those, 11 (73%) had a fasting plasma glucose <7 mm at the last consultation. Overall, fasting plasma glucose decreased from 10.8 (SD ± 5.2) mm at treatment start to 6.2 (±1.7) mm at the last recorded consultation (95% CI: 1.7–7.6, P < 0.01).
Costs of therapy
At 1 year, 34.6% of retained hypertensive patients received HCTZ monotherapy, 3.9% nifedipine monotherapy and 61% a combination of nifedipine + HCTZ (one patient with missing data). Based on the last recorded doses, the average monthly cost to patients for antihypertensive medication was € 1.1 ± 0.9. Most of this was attributable to nifedipine (€ 1.0 ± 0.9). Among the 15 patients with diabetes retained at 1 year, at the last recorded visit, nine received metformin monotherapy, two metformin +glibenclamide combination and three diet and lifestyle measures alone. Based on the last recorded doses, the average monthly cost was €1.2 ± 1.1. Transport costs to the centres were on average €1.1 ± 1.0 for hypertensive patients and €1.1 ± 1.6 for diabetics.
To our knowledge, this small-scale, cluster-randomised trial in rural nurse-led clinics in Cameroon is the first randomised trial in Africa focusing on retention of patients with cardio-vascular disease. The outcomes showed the potential for low-cost interventions to improve 1-year retention rates among outpatients with arterial hypertension and/or type 2 diabetes in a rural setting. Both the use of reminder letters and financial incentives improved retention rates significantly. Further, the introduction of a treatment contract in all groups led to increased overall retention rates compared with earlier epidemiological data from the participating health centres (Figure 2). There were no differences in the responses to treatment among retained patients. This is attributable to the nature of interventions that encouraged adherence to visits, but not drug adherence. Drug adherence might need other or additional interventions.
High rates of loss to follow-up are common to NCCD treatment programmes worldwide, and sub-Saharan Africa is no exception. Most available reports from the region-assessing programmes for hypertension and/or diabetes are from cross-sectional assessments or studies with short follow-up times, but those few studies that do report on follow-up confirm that high patient attrition is a main challenge. Most patients are lost to follow-up during the first few months (Kane et al. 1995; Maro & Lwakatare 1997; Bovet et al. 2008; Katz et al. 2009; Kengne et al. 2009a,b). Our own analysis of routine follow-up data from 2007 to 2009 in the ongoing hypertension and diabetes programme in the region where the study was conducted showed a 1-year retention of 18% (Labhardt et al. 2010). Kengne et al. (2009a) report 14% retention at the 12-month visit in a similar programme conducted in five clinics in the Western Region of Cameroon.
Our results indicate that greater patient and caregiver involvement through a treatment contract in combination with either a small incentive or reminder letters has a marked positive influence on retention. Retention rates with both interventions were double those in the control group. At >60%, the rates achieved are quite respectable for any treatment for chronic disease in any geographic area. Comparison with our earlier data from the same area is even more striking (Figure 2).
When allocating resources to interventions to increase retention rates in CV programmes in low-to-middle-income countries in Sub-Saharan Africa, other treatment programmes of chronic disease, notably HIV/AIDS or tuberculosis, need to be taken into account. The immediate individual as well as public health harm of treatment interruptions in the case of HIV or tuberculosis is usually more severe than for most CV diseases. In these programmes, a more resource-intensive approach to retain patients may therefore be justified. Antiretroviral and tuberculosis treatment programmes may require direct physical patient tracing – an approach widely applied (Krebs et al. 2008; Tweya et al. 2010). However, this approach is quite resource-intensive (Rosen & Ketlhapile 2010) and was therefore not considered for our study. The interventions we studied were designed to suit for hypertension and diabetes patients followed in the environment of task shifting and nurse-led care, which is increasingly being implemented in low-resource settings (Beaglehole et al. 2008). All can be implemented with little additional strain on budgets and without the need for additional staff.
All three groups used a treatment contract as background intervention. Such contracts have the advantages of being virtually without cost, and the approach can be taught in less than a day. Treatment contracts for CV treatments and lifestyle changes have met with mixed successes, but there are few rigorous studies available on the strategy (Bosch-Capblanch et al. 2007). We did not evaluate the effect formally, but there appeared to be an improvement with the contract compared with rates documented earlier in the same area. The invitation-letter intervention was integrated into the health committees, COSAs that are already in place in every village in the area investigated. The only additional cost was transport cost for the member of the health committee in case of longer distances.
The one intervention that added relevant costs to the programme was the financial incentive. It should be noted though that the monetary value of the offered reward was very modest. The average monthly cost to patients for antihypertensive medication was € 1.10 ± 0.9. The average transport cost among these patients was €1 ± 1. Monthly spending of patients with diabetes on antidiabetic medication was €1.2 ± 1.1 and €1.1 ± 1.6 for transport cost. Thus, the reward per patient was slightly more than €1 every 4 months, a quite moderate sum as compared to the transport cost. These calculations also indicate that as in the other interventions, psychological impact was the probable driver of outcomes, rather than the economic benefit.
Another point may be derived from the cost calculations. The cost to patients of drugs was of the same magnitude as expenditures for transport to the healthcare centres. As potent antihypertensive and antidiabetic drugs are available as generics, costs tend to be moderate. Unaffordable drug prices are cited in several studies on barriers to retention (Maro & Lwakatare 1997; Ohene Buabeng et al. 2004; Akpa et al. 2005; Amira & Okubadejo 2007; Magadza et al. 2009). However, our data suggest that transport cost may be at least as much an obstacle as drug prices. In rural areas of low- and middle-income countries, the greatest scope for cost reductions to patients with CV disease may therefore well be through decentralisation of services to rural nurse-led clinics rather than drug subsidies.
For wider implementations of retention programmes in similar settings, we would recommend an intervention with treatment contract and invitation letter, based on the study results and on experiences in the field. The letter system has several appealing advantages over the financial incentive. First, it is less prone to corruption than incentives involving money. Second, the need to send out a reminder based on recorded visits improves the keeping of patient records as nurses need to check these on a regular basis to identify missed visits. Third, the process of delivering the letter involves the health committees in the villages and thus strengthens the collaboration between different local stakeholders in the provision of primary health care.
This study has limitations. First, the calculated sample size of 100 patients per arm with nine clusters per group was not met although the effects from the interventions were significant even in the reduced sample. Second, the cohort of patients with diabetes was very small, and only limited conclusions could be drawn from this group. Third, the study was not designed to assess cost-effectiveness of the interventions, which would be important to considerations of implementation on a larger scale. Fourth, the study only examined visit adherence and not drug adherence. Indeed, both are of high importance to achieve treatment effectiveness. Assessing drug compliance by electronic pill containers in the Seychelles, Bovet et al. (2002) reported a satisfactory compliance by twelve months in only 26% of hypertensive patients. Finally, the promising changes in BP and FPG observed in all three groups should be cautiously interpreted, as a considerable part may represent a regression to the mean that is independent of pharmacologic treatment. This phenomenon has been described in several studies, among others Bovet et al. (2003) in Tanzania.
In summary, this trial showed that low-cost interventions suited to an environment of task shifting and nurse-led care and needing minimal additional resources can significantly improve retention rates in CV disease management in rural Africa. Moreover, if transport expenditures for patients can be reduced by decentralising the management of hypertensive patients and patients with diabetes in rural Africa to NPC clinics, access might be improved without increased strain on healthcare budgets.
We thank the patients and providers for participating in the study; Dr Pelle Stolt for the helpful comments on the manuscript; and Dr Bernard Cerutti for double checking the statistical analyses.