Evaluation of Novartis Access; a Non-communicable Disease (NCD) Access Initiative

Participants, interventions, and outcomes: Eight counties have been non-randomly selected for inclusion in the study. Initially, Novartis Access medicines will be sold solely to the Mission for Essential Drugs and Supplies (MEDS), the main supplier for the large network of private (often faith-based) non-profit health facilities in the country, and also a key supplier for public facilities in several counties. Of the 47 total counties in Kenya, 17 counties were excluded because they do not purchase medicines from MEDS, and an additional 15 counties were excluded that had not purchased at least 10 million KES worth of medicines through MEDS in the previous year. Three counties were excluded due to security concerns, and four additional counties were excluded to eliminate shared borders in the final sample, to minimize the risk of contamination between intervention and control counties. When considering counties with shared borders, those with the lower volume of MEDS purchases in the previous year were excluded.

In Kenya, NCD services are provided in the public and private non-profit sectors at level 2 dispensaries up through level 6 hospitals. Not all counties have a level 6 hospital, and for this reason they are excluded from the study. Counties have on average: five public and private non-profit level 5 county referral hospitals; three public level 4 sub-county hospitals; 15 public and private non-profit level 3 health centers; and 85 public and private non-profit level 2 dispensaries. All 184 eligible level 3-5 facilities will be included in the study, and 200 level 2 dispensaries will be randomly selected for inclusion. For each facility included in the study, one private drug seller identified as the main alternative for that facility will be identified and included. Private drug sellers include pharmacies, drug stores, and dispensing doctors.

A total sample of 800 households will be randomly selected from the eight study counties (400 from Novartis Access counties; 400 from control counties) using a two-stage sampling procedure. In the first stage, 10 enumeration areas (EAs) will be selected in each county with probability proportional to size based on data from the most recent census. Then, in the second stage, 10 eligible households will be randomly selected in each EA and recruited into the study. All households in the EA will be listed in a random order, and enumerators will proceed down the list until 10 eligible households are identified. Households will be eligible if at least one member at least 18 years old has been previously diagnosed and prescribed medicine for NCDs addressed by Novartis Access, including diabetes, hypertension, breast cancer, and asthma. All members of the household who fit that criterion will be recruited for the study. Based on the prevalence of the relevant NCD conditions in Kenya, 20% of all households will meet the inclusion criterion.

In intervention counties, public and private non-profit health facilities will be allowed to purchase subsidized Novartis Access NCD medicines through MEDS. The Novartis Access portfolio includes 15 medicines to treat diabetes, hypertension, breast cancer, and asthma. While the portfolio includes patented medicines usually marketed and sold under the Novartis brand and generics under the Sandoz brand, all medicines sold through the initiative will be packaged with the new Novartis Access brand. The medicines will be sold directly to MEDS at an average cost of 150 Kenyan Shillings (KES), around US$1.50, per monthly dose. Facility-level administrative data on NCD medicine purchases obtained from MEDS will be used to monitor the implementation of the intervention.

The primary outcomes of interest are medicine availability and price at health facilities, and medicine availability, price, and expenditures at patients' households. Availability at facilities is defined as the proportion of Novartis Access medicines and equivalents in stock on the day of data collection. Availability at households is defined as the proportion of prescribed Novartis Access medicines and equivalents in the home on the day of data collection. The price of NCD medicines sold by private for-profit drug sellers is an important secondary outcome. Patient perceptions of NCD medicine access will also be explored. Data will be collected at facilities and households using a structured questionnaire at baseline prior to the implementation of Novartis Access, at midline after one year, and at end line after two years. At each of these interviews, a subsample of facilities and households will also be administered a qualitative interview. Additional data will be collected quarterly from facilities and a randomly selected subsample of households using a structured questionnaire administered over the phone.

The study is powered to detect a 10 percentage point increase in household availability of medicines due to Novartis Access at α = 0.05, assuming an intracluster correlation coefficient of 0.05, 10% loss to follow-up, and a proportion available in the control group of 33%.

Assignment of the intervention: In Kenya, county health offices are responsible for bulk purchasing of medicines for their health facilities, allowing for random allocation of Novartis Access at the county-level. The eight counties selected for inclusion in the trial will be randomized to the intervention or control group using a covariate constrained randomization method to maximize balance on nine demographic and health variables: total population; population density; proportion of the population in urban areas; poverty rate; number of health facilities; physicians per capita; health spending per capita; overall value ordered through MEDS in previous year; and proportion of value ordered through MEDS in previous year by private non-profit versus public health facilities. Allocation will be masked from data collectors. However, it will not be possible to mask the intervention from participants.

Data collection, management, and analysis: Data will be collected from health facilities and households at baseline, midline, and end line using a structured questionnaire. The health facility questionnaire has been adapted from an instrument developed by the World Health Organization and Health Action International and previously used in Kenya, and captures information on medicine availability and medicine price. The household questionnaire includes information on demographics, household assets, and key information on medicines, including whether prescribed NCD medicines are currently at the home, locations where medicines are most often purchased, prices paid, and overall household expenditures on medicines (and on other goods).

A subsample of health facilities and subsample of households will also be administered a qualitative interview during baseline, midline, and end line visits. A purposeful subsample of five level 2 dispensaries and five level 3-5 facilities will be selected in each county (80 facilities in total), and one staff member will be randomly selected for the interview. A purposeful subsample of 10 households in each county (80 households in total) will be selected for qualitative interviews, which will be administered to one individual in the household who also completes the general household survey. Facilities and households will be selected for qualitative interviews to maximize variation in the viewpoints represented. The qualitative instrument has been designed to explore in depth key issues related to medicine access.

Surveillance data will be collected quarterly from all health facilities and a subsample of households. For health facility surveillance, a shortened version of the structured questionnaire with basic information on medicine availability and price will be administered over the phone each month to a rotating one-third sample of health facilities, such that all facilities are surveilled once per quarter. For household surveillance, a random sample of half of all study households will be selected; half of study households will not be surveilled at all, to allow for an analysis of the potential effect of surveillance on household behavior. Those households selected for surveillance will be administered a shortened version of the structured questionnaire with basic information on medicine availability and price over the phone. As with health facilities, a rotating one-third sample of households selected for surveillance will be surveyed by phone each month, such that all surveilled households are surveyed once per quarter. A 5% subsample of surveilled health facilities and households will be visited in person to audit phone responses.

Innovations for Poverty Action (IPA) will be a partner on the study and will manage research fieldwork. IPA specializes in the design and implementation of randomized controlled trials to test policy interventions in developing country settings, and has partnered on several rigorous evaluations in Kenya in recent years. Data will be collected electronically using SurveyCTO software on tablets and managed using Microsoft Access software.

Data on baseline characteristics of health facilities and households will be compared across groups to assess balance. Impacts of Novartis Access will be estimated using intention to treat analysis. Analysis of covariance (ANCOVA) methods will be used to control for potential baseline differences between groups. All analyses will be conducted using Stata statistical software.