South Africa uses a courier network for transporting specimens to public laboratories. After the daily collection of specimens from the facility by the courier, patients not yet attended to are unlikely to receive same-day blood draws, potentially inhibiting access to viral load (VL) testing for HIV patients.
We aimed to design an optimised courier network and assess whether this improves VL testing access.
We optimised the specimen transport network in South Africa for 4046 facilities (November 2019). For facilities with current specimen transport times (
The number of annual VL tests per antiretroviral treatment patient (1.14, standard deviation: 0.02) was higher at facilities that had courier collection after 13:36 (the average latest collection time) than those that had their last collection before 13:36 (1.06, standard deviation: 0.03), even when adjusted for facility size. Through network optimisation, the average time for specimen transport was delayed to 14:35, resulting in a 6% – 13% increase in patient access to blood draws.
Viral load testing access depends on the time of courier collection at healthcare facilities. Simple solutions are frequently overlooked in the quest to improve healthcare. We demonstrate how simply changing specimen transportation timing could markedly improve access to VL testing.
The National Health Laboratory Service (NHLS) is the largest diagnostic pathology service provider in South Africa, providing laboratory and related public health services to over 80% of the population.
In the case of HIV patients specifically, if the patient is unable to return the following day, viral load (VL) testing could be delayed until the next clinical visit or when the patient is required to collect their next antiretroviral treatment (ART) medication – potentially 2–6 months later – due to inadequate cold-storage facilities at PHC facilities for blood samples. Viral load testing is the recommended method for monitoring HIV patients on ART. South Africa, through the NHLS, currently operates a highly centralised national network that conducted more than five million VL tests at 16 laboratories in 2018.
We aimed to design an optimised courier network that could provide expanded service access to all diagnostics available at PHC facilities within the costs of the current system. We then aimed to assess the impact of the optimised courier network on VL testing access.
This project was approved by the University of the Witwatersrand Human Research Ethics Committee (Medical) (HREC) as an additional project within the Integrated Laboratory Data Analysis for Care programme (study approval number: HREC M160978). Additional ethics approval was received from the Center for Global Health, United States Centers for Disease Control and Prevention in Atlanta (study approval number: CGH 2019-224). As this study involved the retrospective analyses of de-identified, aggregated data collected as part of routine care, no patient consent was sought. Although sensitive information was not collected, all data extracted from the NHLS were anonymised, aggregated by facility, and transferred to a secure server. Patients or the public were not involved in the design, conduct, reporting, or dissemination plans of this research.
We extended a previously described geospatial cost model to optimise the specimen transport network in South Africa.
Data from all public sector healthcare facilities were extracted from NHLS’s Corporate Data Warehouse for a 12-month period (January 2018 to December 2018) and used in the model. Extracted data included the number and type of tests (e.g. VL) requested per site, the location of the site, and the laboratory where specimens were sent to be tested. Data from 4046 healthcare facilities from the NHLS Corporate Data Warehouse were included in the optimisation. These 4046 facilities accounted for 98% of the total test volumes that were couriered during the study period.
Facilities and laboratories in the Corporate Data Warehouse were matched to the District Health Information System so that coordinate data from the District Health Information System could be used to locate the facilities and laboratories. Data quality assessments were conducted on the coordinates to ensure accuracy, and missing coordinate data were sourced from internal implementing partner sources, as well as from Google searches.
The optimised system enhanced the functioning of the current specimen routing network by incorporating factors that would improve specimen validity and would be better aligned with patient-centred care (online Supplementary Text 1).
The optimisation of the specimen transport system was conducted using the ArcGIS Pro Network Analyst tool (Environmental Systems Research Institute, Redlands, California, United States) for vehicle routing problems, which optimises a set of transport routes taking into account the expected specimen volumes, distance from the laboratory to the facility, drive times, and other key constraints identified by stakeholders (online Supplementary Text 2).
To assess whether the optimised specimen transport network improved VL access, we first determined whether facilities that historically had earlier courier collection times had lower VL test volumes compared to facilities with later courier collection times. We then quantified the additional number of patients that would be expected to access VL testing due to the expanded access brought about by delayed courier specimen collection times.
To determine the current latest time at which a health facility is serviced by the courier network, we obtained data on current courier routes and the expected time of arrival at each facility by the courier. Unfortunately, arrival time was not well documented and was only available for a subset of facilities in two provinces, Gauteng and Western Cape (
Using STATA software (version 15.0, College Station, Texas, United States), we performed an unadjusted regression analysis to assess the relationship between current specimen transport time at the 356 PHC facilities and VL testing access, defined as the number of annual VL tests performed per patient on ART.
We estimated the time distribution of blood draws for patients at PHC facilities based on data from an unpublished time-in-motion study on wait times conducted at an HIV outpatient facility in Johannesburg in 2014.
A sensitivity analysis assessed the impact of changing the average latest courier time of the current system on the relationship between the time of the last courier collection for the day and the number of VL tests requested per patient on ART. For each time point (representing the average latest time of courier pick-up at a facility), we determined whether there was a statistically significant difference in the average number of VL tests per patient on ART between facilities whose last courier collection for the day occurred before that point versus those that had courier pick-ups after that point. In addition, we assessed how a different assumption regarding the distribution of patient flow (a more uniform distribution) at a PHC facility would change our results in terms of patient access to same-day blood draws.
The optimised network included 1179 courier routes, involved 136 000 km of driving per day, and serviced the 4046 facilities that are part of the NHLS national network at least once per day. The cost to operate the current system is estimated to be $283 488 United States dollars (USD) per week, and the optimised system costs 1.3% less at $279 901 USD per week. Through specimen network optimisation, the average latest time that patients can access blood draw for laboratory services was delayed to 14:48 from 14:15. In addition, with optimisation, an estimated 85% of specimens arrive at the laboratory by 16:00.
In our sample of facilities for which we had complete current specimen transport times, the weighted average latest time an individual can access a blood draw (weighted by the VL volumes at the facility) is currently 13:36. Under half (141/356; 40%) of the PHC facilities had an average latest courier pick-up time before 13:36, while the average latest courier pick-up time at 215/356 (60%) facilities was after 13:36.
The average number of annual VL tests conducted per patient on ART was higher at facilities with courier pick-up after 13:36 (1.14; standard deviation: 0.02) compared to those with courier pick-up before 13:36 (1.06; standard deviation: 0.03). The number of VL tests conducted per patient on ART was higher (by approximately 0.07) among facilities with courier collection after 13:36 compared to facilities with courier collection before 13:36, and the regression coefficient was significant (
Relationship between average latest courier specimen collection time and average viral load tests per antiretroviral treatment patient
Variables | Before 13:36 | After 13:36 |
---|---|---|
Number of facilities | 141 | 215 |
Average VL tests/ART patient | 1.06 | 1.14 |
Standard deviation | 0.03 | 0.02 |
Coefficient | 0.08 |
0.07 |
Standard error | 0.03 |
0.03 |
0.005 |
0.016 |
VL, viral load; ART, antiretroviral treatment.
, Average VL tests/ART patient refers to the average number of VL tests conducted per patient on ART;
, Unadjusted (
, Adjusted (
Data from an unpublished time-and-motion study on wait times show that the timing of blood draws was skewed to the right, with only 17% of patients getting a blood draw after 12:00 (
Hourly distribution of patient blood draws at an outpatient clinic, Johannesburg, South Africa, 2014.
In a sensitivity analysis, facilities that are serviced by a courier earlier in the day are more likely to have a significantly lower number of VL tests conducted per patient on ART (
Relationship between time of courier collection and viral load access at public sector healthcare facilities in Gauteng and Western Cape provinces, South Africa, December 2019.
Through the optimisation of the specimen transportation network, the estimated average time of specimen collection was delayed from 14:15 to 14:48 for the entire network of public facilities (
Improving VL testing access in accordance with national guidelines is critical to attaining South Africa’s goal of high levels of viral suppression among HIV patients to meet the last 95% of the HIV targets of the Joint United Nations Programme on HIV and AIDS.
Optimising the transportation for VL testing is only one mechanism to improve patient access to VL testing. There are alternative strategies that allow for blood to be drawn at any time of the day. If dried specimens (e.g. dried blood spots or dried plasma spots) were used for VL testing, for which they currently are not, specimens could be taken at any time during the day, even after the courier has picked up the specimens for the last time that day.
To our knowledge, this is the first paper to model the relationship between specimen transport collection times and VL access in a national VL programme. This analysis relies on innovative geospatial routing optimisation and primary data on laboratory and facility location matched to programmatic ART data.
This study has several limitations. First, we sampled a subset of facilities in only two, predominately urban, provinces out of the nine provinces in South Africa. Nevertheless, while a more representative sample set may have produced different results, this would primarily have been driven by the current average latest time of specimen collection by couriers, which we varied in the sensitivity analysis. The relationship between specimen transport time and VL access would only become insignificant if the average latest time of specimen transport was later than 14:45, which is unlikely to be the case across the network.
Second, this project was conducted in South Africa with an extensive well-functioning specimen referral network that ensures each facility is serviced at least once per day.
Third, we relied on the patient flow distribution at one PHC facility, where the median time of blood draw was 09:42 in the morning, to quantify the patient benefits of increased access. However, for PHC facilities with later median times of blood draw (e.g., with a more uniform distribution of patients throughout the day), the benefits in terms of increased patient access of delaying the courier pick-up time would be greater.
Fourth, we did not quantify the patient benefits of receiving same-day blood draws. Between 6% and 13% of patients requiring a blood draw for VL testing would not need to return to the PHC facility another day, potentially reducing transport costs/opportunity costs of a missed day of work. Patient scheduling and flow may also improve with later specimen transport as facility staff will be aware that there is more time to send patients to the blood draw queue and can thus book patients in for later times in the day, ultimately reducing the waiting time for patients.
Finally, while this project has primarily addressed PHC facilities that predominantly operate during the week and close at 16:00, future research could look at the value of couriers collecting specimens after hours and on weekends at facilities that operate 24 h and/or on weekends. The benefits in terms of increased patient access to VL testing could even be larger as these facilities could potentially offer phlebotomy services to patients after hours and on weekends, where currently there are none.
Simple solutions are frequently overlooked in our quest to improve healthcare. Delaying specimen pick-up times at PHC facilities by an hour can improve patient access to VL testing at no additional cost. This has implications beyond VL testing and extends to all laboratory diagnostics, especially for specimens whose integrity can be compromised without refrigeration.
The authors would like to thank the National Health Laboratory Service for their valuable input with acquiring the data.
The results from this manuscript have been presented at the Conference on Retroviruses and Opportunistic Infections 2020.
The authors declare that they have no financial or personal relationships that may have inappropriately influenced them in writing this article.
W.S., K.D., L.B., and B.N. conceived the project. S.J.G., B.N. and, T.C. and N.C. acquired and analysed data for the model. S.J.G. and B.N. developed the model. S.J.G. and B.N. interpreted model results. S.J.G. and B.N. wrote the first draft of the manuscript. S.J.G., T.C., N.C, F.O., P.S., K.D., L.B., D.C., W.S. and B.N. read and approved the final manuscript.
This work has been supported by the United States President’s Emergency Plan for AIDS Relief (PEPFAR) through the United States Centers for Disease Control and Prevention under the terms of CoAg 5 NU2GGH001631-04-00: Strengthening the delivery and expanding access to quality laboratory services and enhancing healthcare worker and laboratory safety in the Republic of South Africa under PEPFAR: protocol number 2019-224. The findings and conclusions in the report are those of the authors and do not necessarily represent the official position of the funding agencies.
There are restrictions to the availability of the data used to perform the analysis presented in this manuscript. Requests for the underlying data used in this manuscript need to be directed to the National Health Laboratory Service. (
The views and opinions expressed in this article are those of the authors and do not necessarily reflect the official policy or position of any affiliated agency of the authors.