Patient-Specific Mobile Phone-Generated Reminders and Quality of Hypertension Care in Western Kenya.

To evaluate the impact of clinician-targeted mHealth-generated care suggestions on compliance with hypertension care guidelines in a resource-limited setting. This study was conducted in 10 rural health clinics in Western Kenya that offered hypertension care through nurses and clinical officers. Sites were grouped into intervention and control groups. Intervention group clinicians had patient-specific care suggestions triggered and displayed on a mobile application, mUzima, for their action. Care suggestions were also triggered in the mHealth application for control arm clinicians but were not displayed. Differences in compliance with hypertension care guidelines were evaluated. The study involved 378 patients with hypertension who had care suggestions generated during visits (217 in intervention group and 161 in control group). There was a higher proportion of adherence to hypertension care guidelines in the intervention group compared to the control group (91.1% vs. 85.7%, p=0.014). The random effects model showed significant variability in compliance rates among study clinicians (variance of 0.44, 95% CI: 0.12 -1.62). When displayed care suggestions were rejected by intervention providers, the most common reason given was 'Previously ordered' (58.8%). Clinicians felt that care suggestions improved awareness of hypertension care guidelines. The successful scaled implementation of mUzima with patient specific care suggestions led to higher adherence to hypertension care guidelines and improved quality of hypertension care. Tailormade m-Health applications in resource constrained settings for hypertension care and other chronic non-communicable diseases has the potential to lead to better adherence to care guidelines and quality of care.

Data Cleaning Process for mHealth Log Data to Inform Health Worker Performance.

Log data, captured during use of mobile health (mHealth) applications by health providers, can play an important role in informing nature of user engagement with the application. The log data can also be employed in understanding health provider work patterns and performance. However, given that these logs are raw data, they require robust cleaning and curation if accurate conclusions are to be derived from analyzing them. This paper describes a systematic data cleaning process for mHealth-derived logs based on Broeck's framework, which involves iterative screening, diagnosis, and treatment of the log data. For this study, log data from the demonstrative mUzima mHealth application are used. The employed data cleaning process uncovered data inconsistencies, duplicate logs, missing data within logs that required imputation, among other issues. After the data cleaning process, only 39,229 log records out of the initial 91,432 usage logs (42.9%) could be included in the final dataset suitable for analyses of health provider work patterns. This work highlights the significance of having a systematic data cleaning approach for log data to derive useful information on health provider work patterns and performance.

Feasibility of Cardiac Rehabilitation Models in Kenya.

BACKGROUND: Heart failure (HF), is a leading cause of cardiovascular morbidity and mortality in Sub-Saharan Africa. Cardiac rehabilitation (CR) is known to improve functional capacity and reduce morbidity associated with HF. Although CR is a low-cost intervention, global access and adherence rates to CR remain poor. In regions such as Western Kenya, CR programs do not exist. We sought to establish the feasibility CR for HF in this region by testing adherence to institution and home-based models of CR. METHODS: One hundred participants with New York Heart Association (NYHA) class II and III HF symptoms were prospectively enrolled from a tertiary health facility in Western Kenya. Participants were non-randomly assigned to participate in one of two CR models based on their preference. Institution based cardiac rehabilitation (IBCR) comprised 36 facility-based exercise sessions over a period of 12 weeks. Home based cardiac rehabilitation (HBCR) comprised weekly pedometer guided exercise targets over a period of 12 weeks. An observational arm (OA) receiving usual care was also enrolled. The primary endpoint of CR feasibility was assessed based on study participants to adherence to at least 25% of exercise sessions. Secondary outcomes of change in NYHA symptom class, and six-minute walk time distance (6MWTD) were also evaluated. Data were summarized and analyzed as means (SD) and frequencies. Paired t-tests, Chi Square, Fisher's, and ANOVA tests were used for comparisons. FINDINGS: Mean protocol adherence was greater than 25% in both CR models; 46% ± 18 and 29% ± 11 (P < 0.05) among IBCR and HBCR participants respectively. Improvements by at least one NYHA class were observed among 71%, 41%, and 54%, of IBCR, HBCR and OA participants respectively. 6MWTD increased significantly by a mean of 31 ± 65 m, 40 ± 55 m and 38 ± 71 m in the IBCR, HBCR and OA respectively (P < 0.05). CONCLUSIONS: IBCR and HBCR, are feasible rehabilitation models for HF in Western Kenya. Whereas improvement in functional capacity was observed, effectiveness of CR in this population remains unknown. Future randomized studies evaluating effect size, long term efficacy, and safety of cardiac rehabilitation in low resource settings such as Kenya are recommended.

Leveraging mHealth usage logs to inform health worker performance in a Resource-Limited setting: Case example of mUzima use for a chronic disease program in Western Kenya.

BACKGROUND: Health systems in low- and middle-income countries (LMICs) can be strengthened when quality information on health worker performance is readily available. With increasing adoption of mobile health (mHealth) technologies in LMICs, there is an opportunity to improve work-performance and supportive supervision of workers. The objective of this study was to evaluate usefulness of mHealth usage logs (paradata) to inform health worker performance. METHODOLOGY: This study was conducted at a chronic disease program in Kenya. It involved 23 health providers serving 89 facilities and 24 community-based groups. Study participants, who already used an mHealth application (mUzima) during clinical care, were consented and equipped with an enhanced version of the application that captured usage logs. Three months of log data were used to determine work performance metrics, including: (a) number of patients seen; (b) days worked; (c) work hours; and (d) length of patient encounters. PRINCIPAL FINDINGS: Pearson correlation coefficient for days worked per participant as derived from logs as well as from records in the Electronic Medical Record system showed a strong positive correlation between the two data sources (r(11) = .92, p < .0005), indicating mUzima logs could be relied upon for analyses. Over the study period, only 13 (56.3%) participants used mUzima in 2,497 clinical encounters. 563 (22.5%) of encounters were entered outside of regular work hours, with five health providers working on weekends. On average, 14.5 (range 1-53) patients were seen per day by providers. CONCLUSIONS / SIGNIFICANCE: mHealth-derived usage logs can reliably inform work patterns and augment supervision mechanisms made particularly challenging during the COVID-19 pandemic. Derived metrics highlight variabilities in work performance between providers. Log data also highlight areas of suboptimal use, of the application, such as for retrospective data entry for an application meant for use during the patient encounter to best leverage built-in clinical decision support functionality.

Do diabetes mellitus patients adhere to self-monitoring of blood glucose (SMBG) and is this associated with glycemic control? Experiences from a SMBG program in western Kenya.

AIMS: Among diabetes mellitus (DM) patients with poor glycemic control enrolled into a self-monitoring of blood glucose (SMBG) program in Kenya, to assess the level of SMBG adherence, its associated factors and its relation to glycemic control (defined as HbA1c <7% and/or 2% absolute reduction relative to baseline). METHODS: In this retrospective cohort study, we used routinely collected data of patients enrolled during 2012-2013. We assessed adherence to SMBG by dividing the number of glucose tests performed by the number recommended. A level of ≥ 80% was considered 'good adherence'. Glycemic control was considered as absolute change from baseline of 2%. RESULTS: Of 164 patients (59% female; 76% rural), the proportions with good SMBG adherence were 34%, 17%, 15% and 10% during 0-6, 7-12, 13-18 and 19-24 months into the HGM program respectively. In multivariate analysis, male gender, urban place of residence and payment for glucostrips were associated with poor adherence during 0-12 months. The mean reduction in HbA1c compared to baseline was 1.2%, 1.1%, 0.8% and 0.7% at 6, 12, 18 and 24 months, respectively. We did not find any association between SMBG adherence and glycemic control. CONCLUSIONS: Adherence to SMBG was sub-optimal, especially among those who had to pay for glucostrips. Patient education and provision of free glucostrips are recommended to improve adherence and glycemic control.