The Computer-based Drug and Alcohol Training Assessment in Kenya (eDATA K)

The purpose of this project is to answer the challenge of integration of screening and core packages of Mental Health (MH) and Substance Use Disorder (SUD) services into routine primary health care (PHC) through innovative e-learning technologies. The training of health professionals in Kenya (and other LMIC) has very limited content on addressing alcohol, tobacco and other substance use disorder in PHC. The investigators aim to help address this problems by adapting NextGenU training model for capacity-building of health care providers in LMIC to provide SUD screening and core services in PHC. NextGenU.org draws on already-existing, free, expert-created competencies and computer-based learning resources from accredited sources to assemble courses certified by its network of partners. The computer-based didactic learning experience is complemented with 1) peer-to-peer interactions in a local and global community of learners simultaneously studying this topic, and 2) the use of local qualified mentors, which can supplement the training taking into account local variations in organizational structures, culture, tradition and beliefs.

The intervention evaluated through this program of research includes training lay health worker or facility support staff to screen patients presenting to the facility. Those support staff then refer those who are at moderate or high risk from tobacco, alcohol or other substance use to trained PHC clinician for brief intervention (BI) related to any of the substances they use at risky level. The training intervention includes teaching the clinicians to deliver BI, and identify and manage co-morbidities or complications. The program of research around that intervention includes (1) a developmental evaluation of the on-line training through a pilot with students and clinical faculties in educational institutions, followed by the assessment of the training of already practicing health care providers in the field; (2) pre and post training assessment of clinicians and non-clinicians knowledge, attitudes and skills (including assessing their level of stigma toward those who use psychoactive substances); (3) RCTs to assess the impact of the BI on alcohol consumption and other health and social variables of interest in public and private facilities; and (4) the assessment of the impact of quality improvement training on sustainability of the screening and BI. This trial registration is only for the RCTs.

Two independently powered RCTs, one in private facilities and one in public facilities, will be run. The rational to run two independently powered RCT is that other studies of PHC workers training for depression in LMIC have shown that the intervention implemented in public facilities was successful at improving patients' outcomes, while it was not in private facilities. However, since a large proportion of the population uses private facilities, the investigators did not want to exclude private facilities from participating.

The RCT with public institutions takes place in two rural counties, Machakos and Makueni, chosen as they are very similar in terms of geography and socio-demographic characteristics, and are pretty typical of rural counties in Kenya and other LMIC. They both have a population of over 700,000, with a moderate burden of substance abuse for Kenya.

The RCT with private health care institutions (CliniX and Shalom) will takes place in the Nairobi metropolitan area. The Shalom facility is technically in the County of Machakos, very close to Nairobi city limit and serving a very similar population to that of the CliniX facilities (urban middle class Kenyans who can afford private health care consultations out of pocket or who have insurance through their workplace). The RCTs will be complemented with qualitative methodologies to monitor other aspects besides the clinical effectiveness of the intervention. On-going results will be shared with stakeholders, and may result in adapted implementation of the training as the study progresses, as per the developmental evaluation methodology. The context and rationale of any changes will be documented as part of the research process, and are expected to inform the development of the innovation, including scalability.

The methodology for the RCTs is based on the methodology followed by Flemming et al in their studies on BI for alcohol. These highly-ranked studies have been included in the latest version of the Cochrane meta-analysis on brief alcohol intervention.

Eligible sites for the RCTs includes those which are offering PHC, are willing and able to participate in the training program (access to electricity, internet through cell phone networks, staff expressing interest in being trained), have a sufficient number of clinicians and support staff who have completed the training to screen and deliver BI, and are amenable to follow the research protocol. The investigators aim to enlist in the training at least 4 public facilities per district, and 4 private facilities, as there are uncertainties over how many clinics will have their staff complete the training.

At each clinic, a community health worker (in public facilities) or complementary staff (in the private clinics) will offer patients presenting for consultation who are aged 18 years or older to be screened for healthy lifestyle and to receive feedback on their results. The person performing the screening will obtain consent specifically to participate in the screening. To mask the topic of the trial at this early stage and avoid potential stigmatization of participants, the study will be presented as one about general lifestyles (as was done in the Flemming et al. studies). The screening will include self-reported weight and height, (for body mass index - BMI), level of physical activity, and the Alcohol, Smoking and Substance Involvement Screening Test (ASSIST). The inclusion of the BMI and level of physical activity is to mask the purpose of the intervention, by decreasing the focus on substance use. The lay health worker will provide the results of the BMI, physical activity and the ASSIST and will offer a leaflets with further explanation about those lifestyles. The screener will send all of those who screen to moderate risks or above from alcohol use to the research assistant (RA). The RA will offer patients to participate in the RCTs, assess eligibility, obtain consent, collect baseline data and demographic, and randomize patients to either BI or no further intervention (the patients will have already received their screening results). The RA will explain that it is a study about lifestyles, without disclosing that he is asking them to participate because of their level of risk due to alcohol. Those eligible will be randomized to either the BI or the control group separately for men and women in each site using a computer-generated allocation method. All practitioners will have both control and intervention subjects in their practice.

Those in the BI arm will be asked by the research staff to show and discussed with the clinicians their screening results. The clinician will deliver the BI for alcohol use reaching moderate or high risk level from the ASSIST as well as advice to stop the consumption of harmful levels of any substances reported by the patient. The BI should include discussion of risky level of alcohol consumption, associated adverse effects, and followed the adapted ASSIST BI steps for Kenya, which include asking the patient to come back for follow-up about a month after the BI. Clinicians will record their baseline and follow-up clinical assessment and management through clinical follow-up forms. The total BI should be around 15 minutes, with a follow-up visit of about the same length. Any further follow-up will be at the discretion of the clinician, as part of on-going PHC.

To blind the subject to the control or the intervention allocation, they will be told that the trial focuses on healthy lifestyles including weight, substance use, and exercise. The clinician and complementary health workers will not be told which of their patients are randomized in the control or intervention arm. This methodology ensures appropriate concealment and should prevent contamination. The participants in both the control and the intervention arms will be contacted for follow-up interviews at one, three and six months to assess outcomes. Sub-samples of both control and intervention subjects will be selected for participation in focus groups to seek input into their experience of using the new knowledge and skills in practice, and how it might be improved in the Kenyan context.

Those patients who are ineligible on the basis of increased acute risk will be asked to discuss their ASSIST results with their PHC provider.

The primary outcome variable of interest is difference in mean alcohol consumption (previous 7 and 14 day use measured in g/week) between subjects in the intervention group and the control group, and collected through a timeline follow-back assessment. A statistical test frequently used in this type of study is an independent samples t-test. We used the SPSS SamplePower calculator to estimate the sample size needed to achieve 80% power, with a two-tailed level of significance (alpha) of 0.05. We use the number of grams of alcohol consumed per week obtained from the 1997 Flemming trial: a mean (SD) of 137.7 (135.7) g/wk in the intervention group, and 185.5 (155.2) in the control group. The sample size needed is then n=146 in each arm. The average consumption in the Flemming trial was on the lower side, with a difference in mean reduction very similar to the other studies with adequate allocation concealment included in the Cochrane review.

The assumptions used to estimate the recruitment period needed to reach that sample size are as follows: Using a conservative estimate of number of visits per day from those 18 years old and over (30 visits for each public clinic and 55 visits for private clinic), with 10% a prevalence of moderately risky levels of alcohol use, based on a study conducted in similar PHC centres in Kenya using the ASSIST; subtracting returning patients (an average of 30%). An enrollment of 80% was assumed. This is a conservative estimate, since enrollment has reached about 95% in previous studies of drug and alcohol use or mental illness, conducted by AMHF. Based on these assumptions, about 570 patients in the public clinics and 520 in the private clinics should be enrolled in three months. Loss to follow-up is estimated at 10% at one month, another 10% at three months, and another 10% at six months. At six months of follow-up it is expected to have approximately 370 recruited patients in public clinics and 340 patients in private clinics. This would result in sufficient power to detect the expected decrease in consumption separately in the public and the private sector. If for any unforeseen reason, the estimated needed sample size is not reached in three months, the enrollment period will be extended, and still have the capacity to do a 6 months follow-up on all the recruited subjects.

The lower average effect size obtained by including all studies in the Cochrane review, despite a lack of adequate concealment or other methodological issues, is a reduction of about 33g/week. Using a two-tailed t-test, a consumption by the intervention group of 117g/week versus a consumption of 150 g/week by the control group, a SD of 120 for a 95% CI of (-53.66,-12.34) and an intention to treat sample at initial enrollment of 520 (260 in each arm), the study will have a power of 87.9% to yield a statistically significant result.

The effect size used is the average across men and women. The study is not powered to assess the separate effect on each gender, but sub-analyses estimating the effect of gender as a covariate can be done. Potential confounders and effect modifiers will be tested in multiple linear and logistic regression models depending on the distribution of the outcome variable. The longitudinal data across our four proposed time points will be analyzed using a general linear model, repeated measures technique, in order to understand the effect of variations due to groups (i.e., a between-groups effect) and due to follow-up time (i.e., a within-groups effect). An interaction term between groups and follow-up time will also be fitted. Missing data will be handled through multiple imputation techniques as necessary.