SMS for Life: a pilot project to improve anti-malarial drug supply management in rural Tanzania using standard technology

Objectives

The objectives of the SMS for Life pilot were three-fold: (1) to demonstrate that visibility of weekly stock levels of key anti-malarial medicines at the health facility level will promote action to eliminate and/or reduce stock-outs (2) to demonstrate that a state-of-the-art data gathering infrastructure can be made available via simple tools such as SMS and mobile telephones in remote locations in sub-Saharan Africa (3) to demonstrate the effectiveness of a public-private partnership model.

Location

Of the 131 districts in Tanzania, three rural districts (Lindi Rural, Ulanga and Kigoma Rural) were selected by the NMCP for inclusion in the pilot, covering a total population of 1.2 million. The selected districts met all four criteria for inclusion. First, the districts were to differ in terms of level of health service delivery and access, with the aim of providing a broadly representative sample of the entire country. Lindi Rural is an 'average' district. Ulanga is a challenging district in terms of staff shortages, skill level and remote location. Kigoma Rural also presents problems, due to its large geographic size and long distances between the Zonal Store and remote health facilities. Second, the districts were all to be in different regions of the country, and supplied by different Zonal Stores. Third, all districts were to be malaria endemic with malaria the most common cause of death. Fourth, selected districts were not to be involved in other pilot projects.

Lindi Rural, Ulanga and Kigoma Rural districts included 48, 30 and 51 health facilities, respectively i.e. 129 health facilities in total. The Lindi Rural and Kigoma Rural districts operate anti-malarial supply using a 'pull' system via ILS. The Ulanga district is undergoing a transition from a 'push' system to the 'pull' system.

Duration and scope of the SMS for Life pilot

The pilot study was 21 weeks in duration. This period was chosen because it covered two quarterly order cycles and five monthly delivery cycles. Data collection started on 1^st October 2009 and ended on 25^th February 2010.

The system covered stocks of artemether-lumefantrine (AL, Coartem^®, Novartis Pharma AG, Basel, Switzerland) and injectable quinine (provided by multiple manufacturers). Stocks of four different dosage packs of AL were included: 'yellow' packs used for babies weighing 5 kg to < 15 kg, 'blue' packs for children weighing 15 kg to < 25 kg, 'red' packs for children weighing 25 kg to < 35 kg and 'green' packs for children weighing 35 kg or more and for adults.

The SMS for Life system

The system consists of two components: an SMS management tool and a web-based reporting tool.

SMS management tool (Figure 1)

The SMS application stores a single registered mobile telephone number for one healthcare worker at each health facility. Once a week, a stock request is sent by SMS to each of these telephone numbers. Stock messages are sent back in reply using a free short code number at zero cost to the healthcare worker i.e. telephones do not need to be in credit to reply. A standard message format is used to capture stock quantities of AL and quinine, and formatting errors are handled through follow-up SMS messages to the facility.

Step 1

A personal mobile telephone number for one healthcare worker at each health facility in the three pilot districts was obtained during training sessions and registered with the SMS application. Only stock count messages from registered personal mobile telephone numbers are accepted.

Step 2

Every Thursday at 14:00 an SMS message is sent to all registered health facility workers requesting stock counts.

Step 3

Full boxes of AL in the storeroom of each facility are counted, and individual quinine injectable vials are counted in the storeroom and dispensary (the difference in accounting methodologies was at the request of the NMCP).

Step 4

An SMS message is composed by the health facility worker, comprising a code for each type of medicine and the quantity, following an agreed format.

Step 5

The heath facility worker either replies to the stock request SMS or sends a new SMS using the free short code number. If the message is sent in an incorrect format, the system automatically informs the sender. After three unsuccessful attempts, the district management is informed and asked to intervene.

Step 6

The SMS system sends an automatic reminder to all health facilities that have not replied by Friday at 14:00.

Step 7

The SMS system credits the healthcare worker's mobile telephone with a fixed amount of money (1000-1500 TZS, depending on the district) for personal use if the stock count reply is received before 17:00 on Friday. Late SMS replies are accepted until 13:00 on the following Thursday, but no credit is applied to mobile telephones for late replies.

Step 8

The system provides a weekly status report to the DMO indicating (a) which health facilities did not provide a stock count and (b) which health facilities have a stock-out.

District-level management

The DMO appointed one person in the district to redistribute medicines in response to stock-outs identified by the SMS for Life system. Redistribution could be undertaken by telephoning health facilities with stock-outs to inform them of excess stock in a neighboring health facility, or by contacting the Malarial Focal Person in the district to request that they move stock from a health facility with a high stock level to a neighboring facility.

Participant training

Monitoring and evaluation

Weekly stock reports, stock-out statistics, error rates, deliveries and system access were monitored daily online during the 21-week pilot study. Surveillance visits were undertaken for 116/129 health facilities (90.0%) at least once to validate the accuracy of stock count data provided by health facility workers.

District management team members were interviewed towards the end of the pilot study to assess stock movement during the study, obtain feedback on use and ease of access to the data system and on use of the registration/de-registration function for health facility mobile telephone numbers, seek views on training and training materials, and elicit opinions on the SMS for Life project versus other stock management practices and the potential for future implementation of the scheme. Throughout the project, information on every order and delivery of AL or quinine injectable from Zonal Stores was collected.

Project partnership and contributions

The project partnership had a fixed-term commitment of less than one year, with no centralized budget, formal contract or memorandum of understanding. The Tanzanian Ministry of Health and Social Welfare, The Roll Back Malaria Partnership, Novartis Pharma AG, Vodafone and IBM took part in the pilot project. Each partner funded their own activities.

The NMCP in Tanzania, operating as part of the Ministry of Health and Social Welfare, was the owner and main user of the SMS for Life pilot and coordinated all project activities in the country i.e. planning, implementation and evaluation, including provision of a project leader and vehicles with drivers. The Roll Back Malaria Partnership provided project oversight, including the work of the steering committee, and led advocacy activities. Novartis initiated and led the pilot, defining the solution, sourcing partners, establishing the steering committee, and providing the necessary resources and funding (e.g. to support health professional training). Vodafone and its partner, Matssoft, supported the design, funding and development of the system application and the implementation of the technical solution, and funded all technical operational costs of the pilot. IBM supplied management resource support to the project and provided an on-line collaboration tool 'Lotus Live', which allowed all the project partners to coordinate their inputs across company networks.

Data collection

During the 21-week study, the average response rate to SMS requests for stock count data was 95%. The response rate did not fall below 93% at any point (Figure 2). The proportion of late replies (i.e. after 17:00 on Friday) was low, averaging 3% overall. The rate of responses, and the proportion of late responses, did not vary markedly during the pilot, other than after the request sent on 14^th January 2010 when there was a national problem with connectivity on one mobile telephone network (Figure 2). The highest response rate was in Lindi Rural (99%), compared to 93% in Ulanga and 94% in Kigoma Rural, which is likely to have been the result of disciplinary action in the Lindi Rural district, consisting of warning letters and interviews at the district office for non-compliant health facility staff. Across all three districts, feedback from district management and data from questionnaires completed by health facility workers indicated that the financial incentive of airtime credit was an important contributor to the high response rate.

The error rate for composition of SMS responses was low, averaging 7.5% throughout the study (Figure 2). In Lindi Rural, 100% of error SMS responses were corrected, and although data on corrected rate were not routinely collected the fact that the accepted response rate did not fall below 93% at any point confirms that even incorrect messages from the other two districts were usually corrected.

Stock counting, as assessed by surveillance visits to 116 of the 129 health facilities in the three districts, showed a data accuracy of 94% i.e. the most recent stock message matched the inventory inspected at the health facility.

System usage

The central NMCP log-in was activated on average once a day. The central Medical Stores Department and the Zonal Stores in the three districts virtually never accessed the system. At the district level, the weekly emails sent by the SMS for Life system were read by at least one team member in the district management team of each district every week during the pilot, with the exception of a single email to the Kigoma team. System usage in the Lindi Rural district decreased as stock-outs were eliminated after week 8, declining from 45 log-ins during October 2009 to 13 log-ins during February 2010. In Ulanga, log-ins increased (35 log-ins during October-December, rising to 70 log-ins during the last 6 weeks of the project) after the Clinical Officer in the District Medical Office was given a Blackberry and more prescriptive input from the SMS for Life project team. In the third district, Kigoma Rural, access to the system was low in the early phase (33 times in October-December) but increased to 28 times in the last 6 weeks after the District Pharmacist and Malaria Focal Person were each given a Blackberrry device to access stock count data.

Anti-malarial stock levels

At the start of the pilot (week 1), 78% of health facilities had no stock of one or more of the four different AL dosage packs or of quinine injectable. By the end of the pilot (week 21), this proportion had fallen to 26%. The reduction in stock-outs was largely related to improvements in stocks of AL, since the proportion of health facilities with stock-outs of quinine at the start of the study was lower (18% compared to 77% of facilities with a stock-out of AL) (Figure 3). Stock-outs of all dosages of AL showed a progressive decline over the first two months of the pilot, with a gradual increase from the middle of December to the second half of January, reflecting the ILS delivery schedule. By the end of the pilot, stocks-out of AL blue, green and yellow were almost eradicated but a fifth of health facilities still had no AL red, almost entirely due to continuing stock-outs in the Kigoma Rural district (Figure 4a). Over 80% of facilities held stocks of quinine injectable at baseline, which increased to more than 95% by the end of the pilot (Figure 4b).

Over the same period, total AL stock across the three districts increased by 64% from 2,696 boxes at week 1 to 4,411 boxes at week 21, while the number of quinine vials increased by 36% from 12,536 to 16,981 (36%). Stock levels showed a small increase for all AL dosages by week 21, with similar levels of AL blue, green and yellow, but stocks of AL red remained lower than for other dosages, again primarily due to the Kigoma Rural district (Figure 5a). Quinine injectable stock levels also showed a small increase during the pilot (Figure 5b).

There were marked differences between the three districts in terms of achievement of full stocking and in stock levels, for a variety of reasons. The Lindi Rural district was the most successful in managing stock levels, eliminating stock-outs for all five categories of medicine by week 8 and maintaining stocks of all three anti-malarials at almost all health facilities thereafter. Two key factors contributed. First, after receiving the first set of stock count data, the district management team made an emergency order to the Zonal Store. This delivery was distributed to health facilities according to priority based on their urgency of demand during weeks 2, 3 and 4, thereby eradicating most stock-outs. Second, when a health facility reported having only one box of any AL dosage pack, the district pharmacist either issued further stock or moved stock from a neighboring health facility in a pre-emptive manner. In the Ulanga district, the rate of stock-outs at week 1 was high (87% of health facilities), largely because no blue dosage packs of AL had been delivered to the district for almost a year. Also, Ulanga was transitioning from the 'push' system to ILS delivery during the pilot. As a result, deliveries were delayed and there were discrepancies between stock orders and the item delivered, for example with no blue AL dosage packs included and only very small quantities of other AL dosages. Furthermore, an emergency delivery was not received. Following two ILS deliveries, the second of which included blue AL dosage packs, 78% of all health facilities in Ulanga became fully stocked by week 21. The proportion of health facilities with no quinine injectable, however, increased from 3% at week 1 to 7% at week 21. In the third district, Kigoma Rural, almost all health facilities (93%) had a stock-out of at least one type of anti-malarial at week 1, and 36% were out of stock of all five products. There was an ongoing shortage of red AL dosage packs until the end of January 2010, with 42% of health facilities still having no red packs at the end of the study. Over 90% of health facilities, however, had stocks of all other products by week 21. The district relied only on regular ILS deliveries. Following the two ILS deliveries that were received during the pilot, the district management took 3-4 weeks to distribute medicines from the first delivery to all health facilities and after delivery of a complete ILS order in late December, including red AL dosage packs, stock counts of red packs only rose from 21 January onwards. Several factors contributed to outcomes in the Kigoma Rural district. The ILS delivery quantity for red dosage packs of AL was only sufficient to prevent stock-outs for three weeks, such that stock-outs were inevitable. Second, when red dosage packs were delivered they were distributed unevenly between health facilities, with some facilities receiving none, and no active redistribution was undertaken subsequently. Lastly, no emergency orders were submitted from the Kigoma Rural district despite severe stock shortages for the majority of the pilot.