Am. J. Trop. Med. Hyg., 92(6), 2015, pp.

1271–1279
doi:10.4269/ajtmh.14-0413
Copyright © 2015 by The American Society of Tropical Medicine and Hygiene

Use of eCompliance, an Innovative Biometric System for Monitoring

of Tuberculosis Treatment in Rural Uganda
Sarah Jane Snidal, Genevieve Barnard, Emmanuel Atuhairwe, and Yanis Ben Amor*
School of International and Public Affairs, Columbia University, New York; Millennium Villages Project, Mbarara, Uganda;
Global Health Corps; The Earth Institute, Columbia University, New York

Abstract. Directly observed therapy short-course (DOTS) requires direct observation of tuberculosis (TB) patients
and manual recording of doses taken. Programmatically, manual tracking is both time-consuming and prone to human
error. Our project in western Uganda assessed the impact on TB treatment outcomes of a comprehensive patient support
program including eCompliance, a biometric medical record device, with the aim of increasing TB patient retention.
Through an observational study of 142 patients, DOTS outcomes of patients in the intervention group were compared
with two control groups. Descriptive statistical comparisons, case-cohort analysis, and difference in change over time
were used to assess the impact. Intervention patients had a higher cure rate than all other patients (55.6% versus 28.3%
[P < 0.01]) and the odds of having a “cured” outcome were 3.17 higher (P < 0.05). The intervention group had a
statistically significantly lower odds of having a negative outcome (0% versus.17% [P < 0.01]) than patients from the
control groups. Additionally, the intervention group had a lost to follow-up rate lower than all other groups (0% versus
7%) that was trending on significant. In resource-limited settings, implementing comprehensive DOTS including
eCompliance may reduce the occurrence of negative DOTS outcomes for patients.

BACKGROUND datory daily supervision of drug intake, usually at the clinic,

followed by a 4-month continuation phase, with recom-
Despite over 20 years of intensified global control efforts, mended daily supervision. The DOTS programs, as currently
tuberculosis (TB) still causes approximately 1.4 million deaths designed and implemented, place an important emphasis on
every year.1 One-third of the world’s population is currently monitoring therapy, but they cannot ensure that treatment
infected with the TB bacteria, leading to an estimated annual adherence is fully maintained. Programmatically, manual
incidence of 8.8 million active cases of the disease world- tracking is time-consuming. It is nearly impossible for health
wide. As with many other infectious diseases, the burden of workers to ensure that all patients have been seen and prop-
TB falls disproportionately on the developing world,2 espe- erly recorded, or to verify that records are correct, especially
cially where the incidence of human immunodeficiency virus in high-burden, high-volume settings where hundreds of
(HIV) co-infection is high.3,4 Tuberculosis is still mostly cur- patients need daily supervision.9 Moreover, paper-based
able when the appropriate course of drug therapy is com- records delay the reporting of important patient and disease
pleted, but the full course of antibiotics lasts 6 months, and information, which limits potential implementation of inter-
failure to adhere to complete treatment can lead not only ventions and policies that could greatly improve delivery of
to the return of symptoms, but also to drug resistance.5 Glob- and adherence to TB care.
ally, the reported rate of TB patients lost to follow-up during A suitable solution for both patients and healthcare
TB treatment was > 12% in 2011.1 Consequently, the inci- providers is to implement a biometric system† known as
dence of drug-resistant TB is on the rise: 310,000 cases of “eCompliance” integrated into a comprehensive patient sup-
TB resistant to the two most potent antibiotics, termed port package, to reduce the constraints of manual tracking.10
multidrug-resistant TB (MDR-TB), were reported in 2011.6 The eCompliance system simplifies the monitoring of all
As the incidence of MDR-TB grows, it is increasingly urgent patients on treatment by scanning the patient’s fingerprint
to ensure case detection and adequate treatment completion. at a netbook computer terminal during every visit, whether
To compound this issue, incomplete treatment of TB leads to in the home or at a clinic. The eCompliance system stream-
MDR-TB, but incomplete treatment of MDR-TB can lead lines the TB compliance recording process and ensures that
to extensively drug resistant TB (XDR-TB), a form of TB that health workers see all patients as scheduled by providing
has a much higher mortality rate and is very costly to treat.7 them with immediate access to accurate patient visit logs,
Preventing drug-resistant TB through strict monitoring is removing any guesswork, human error, or time delay caused
therefore critical as mortality in drug-resistant cases is higher, by the upkeep of manual DOTS recording. With this bio-
and treatment is more difficult, lengthier, costlier, and associ- metric system, each patient is registered by adding his or her
ated with more severe side effects for the patients. fingerprints into the secure database. During all subsequent
Directly observed therapy short-course (DOTS)—where a visits for DOTS, patient visits are logged by simply recording
health worker watches a patient take the dose of TB antibi- the patient’s fingerprint again. Within seconds, the health
otics and then manually records the intake on an individual workers can easily view the patients that have not been
patient card—is the World Health Organization (WHO) stan- observed taking a scheduled dose and follow-up accordingly.
dard approach to ensuring retention of patients in treatment The system allows health workers to adequately follow up
programs.8 A DOTS treatment protocol is usually separated on all patients. Those who discontinue treatment can easily
into two distinct phases: a 2-month intensive phase, with man-

*Address correspondence to Yanis Ben Amor, 475 Riverside Drive, †A means of identifying humans through unique traits such as
Suite 520, New York, NY 100115. E-mail: [email protected] fingerprints.

1271
1272 SNIDAL AND OTHERS

be identified and actions can be taken by health workers to Worker (CHW). In addition to the high rate of patients lost to
ensure TB treatment is completed. follow-up, many patients’ adherence was not tracked using the
In this way, the system has clear benefits for both health official patient medical record card issued by the national TB
workers based in the field and in supervisory or monitoring control program. Instead, health workers often used paper note-
positions. For those in the field, it can help them to ensure books to record TB patient information. This inconsistency in
that they are seeing all of their patients, and verifying that record keeping further complicated streamlining of services and
they are seeing the correct patients. Additionally, it can help ultimately, aggregation of data for reporting purposes.
supervisors to ensure the efficacy of the work that their teams
are doing, particularly as the biometric requirement ensures METHODS
that all information reported by field workers is true, and that
all patient visits are recorded at the actual time of visit— Materials. Three components were required to conduct the
thereby limiting human error. It can also help both levels of pilot project at the Ruhiira MVP site: a notebook computer
health workers to review if there are any patterns that may be (our choice: Asus Eee PC model, Taipei, Taiwan), the
indicative of patients that require additional support. eCompliance software (developed by Operation ASHA in
The eCompliance system was developed in India by the tech- India, accessible at http://www.opasha.org/our-work/edots-
nical team at Operation ASHA (http://www.opasha.org/) and innovation-and-health), and a fingerprint scanner (our choice:
has been tested in urban slums since 2010. Results in India show Digital Persona U.are.U 4500, Redwood City, CA). The indi-
a significant decrease in the rate of patients lost to follow-up. vidual cost of the materials was US $200, US $0, and US $70,
The Indian Revised National TB Control Program (RNTCP) respectively. Additional costs only included a one-time train-
suggests average lost to follow-up rates of 7% in the country, ing/implementation cost ($500 for a 2-day training). There
and the Operation ASHA centers using the eCompliance sys- were no additional costs associated with airtime (the short mes-
tem found that the lost to follow-up rate decreased to < 3%.10 sage system [SMS] function was not active in our machines),
The large gains by the Operation ASHA team have been deter- hardware maintenance (we did not incur this cost for the dura-
mined to be largely a result of the system supporting concerns tion of the pilot) and software support was provided free of
related to potential human error. In settings with large TB charge by Operation ASHA.
caseloads or where health workers are handling many different eCompliance software. The eCompliance software was found
responsibilities besides TB management, it can be difficult for to be simple and straightforward for those that are not familiar
health workers to ensure that they are providing each TB with computers. At the Ruhiira cluster, it was operated by
patient with the support that they need to adhere to treatment. CHWs in the homes of each patient. Our description in this
The eCompliance system supports health workers by providing piece therefore assumes the CHWs as end-users. However,
them with active reminders for each patient and comprehen- eCompliance can also be implemented at the clinic level for
sive, accurate records at the touch of button/scan of a finger- trained healthcare workers.
print. In this way, it does not passively provide records, but When the eCompliance system turns on, it automatically
instead plays an active role in giving health care providers and opens to the eCompliance Welcome Screen (Figure 1A). The
teams necessary support to ensure good outcomes. software was developed to allow several types of visitors
The aim of our pilot project was to assess the transferability (Figure 1B). Visitors who are already registered can login by
of the eCompliance system from an urban setting, where it scanning their fingerprint. A patient login only allows regis-
had already been shown successful by Operation ASHA, to a tering a supervised dose. The CHW (or counselor) and pro-
rural setting in Uganda; we assessed whether eCompliance, gram manager login allow full functionality of the software.
when integrated into a comprehensive patient support pack- The CHWs can select the “New Visitor” tab (Figure 1A)
age, can provide substantive improvements to TB DOTS pro- to register new visitors (including patients), edit an existing
grams. The ultimate goal of these improvements is to increase patient’s status using the “Edit Visitors” tab (e.g., a CHW
TB patient retention, which should reduce drug resistance and can change the status of a patient from “active” to “lost
death caused by inadequate TB treatment. to follow-up”), view visitor logs and patients’ missed doses
We implemented the eCompliance system at the Millen- (Figure 1C), send reports, and synchronize multiple eCompliance
nium Villages Project (MVP - www.millenniumvillages.org) system records. However, the latter two functions were not
cluster in Ruhiira, Uganda. The Ruhiira cluster lies in the used during the pilot project because of the small scale of the
southwestern region of Uganda in Isingiro district, about project and to simplify the support provided by eCompliance
40 kilometers from the nearest city, Mbarara. The cluster is to the CHWs. Registering a new patient takes < 10 minutes,
made up of eight villages spread out over several hundred whereas registering a supervised dose takes less than a minute.
square kilometers. We selected Ruhiira to pilot eCompliance Maintenance for both the hardware and the software was
because of its previously reported high rates of TB patients conducted by the MVP electronic Health (eHealth) team.
who were lost to follow-up or died in 2010 (19% and 15.3%, Virus protection software was installed on machines prior to
respectively). The site also has a high TB caseload, with over implementation, and CHWs were instructed to only use the
50 patients diagnosed a year in a population of about 50,000. netbooks for eCompliance work.
Before the implementation of eCompliance at the site, health Ethical approval. This study was reviewed and approved by
workers followed Uganda national TB guidelines implement- the Institutional Review Board at Columbia University (Proto-
ing community-based DOTS (CB-DOTS): after diagnosis and col no.: IRB-AAAJ9911) and by Mbarara University of Science
treatment initiation at a clinic, a family member assigned to and Technology (Protocol no.: MUIRC 1/7). Approval was
observe and record doses performed DOTS daily and manually obtained for adult patients (> 18 years of age). Participants
at the household level. Family supervision was complemented provided verbal informed consent to participate. Written con-
with monthly supervision at home by a Community Health sent was not obtained to prevent selection of participants on
 USE OF TB eCOMPLIANCE IN UGANDA 1273

Figure 1. Accessing eCompliance functions. (A) A registered visitor can instantly log a return visit on the Welcome Screen by simply scanning
one finger twice, which takes less than a minute. Community health workers (CHWs) can access necessary functions of the eCompliance system by
selecting one of the tabs at the top of the screen. (B) From the “New Visitor” page, the CHW can register different categories of visitors. Selecting
the “Patient” option requires the most intensive registration process but will normally take < 10 minutes during the first visit. This process will
establish a visit schedule for each patient that must be maintained (though can be edited) or the system will mark a missed dose in the log. The
other registration processes require information according to the relevant task of the visitor. “Other visitor” is the most basic setting; it allows no
access and affects no data recording—it is typically used for demonstrations. (C) From the “View Visitor” page, a health worker can access a list of
patients to visit, or verify those already visited. Because the interface used in India was both in Hindi and English, the same interface was simply
kept for Uganda where English is the main language.

the basis of literacy. Participants’ consent was recorded the Their patient load was determined by the number of TB cases
following way: the consent form was read by a nurse or com- at their respective clinics, and occasionally a patient was
munity health worker in the local language to each individual switched to a CHW other than the one at their primary clinic
participant who was then asked to circle “yes” or “no” at the if logistically advantageous (e.g., caused by necessary time
bottom of the form based on their intent to participate or not, constraints, proximity to a different observation route, etc.).
and an index fingerprint was affixed next to their answer. The Trainings. The eCompliance system was introduced to the
nurse or CHW obtaining consent subsequently signed and Ruhiira cluster through a series of classroom training sessions
dated each form. Only patients circling “yes” were enrolled over 2 days and patient enrollments at the household level
in the study. This consent procedure was approved by the supervised by the research team. The training component
ethics committee. included four sessions. During the first session, the research
CHW selection and patient load. Three CHWs from different team introduced the project and the objectives to the CHWs,
clinics in the MVP cluster (Kabuyanda, Ntungu, and Ruhiira) instructing them on individual tasks required by the project
were selected to oversee the eCompliance system. The CHWs and the patient observation schedule: daily for patients in the
were chosen based on their catchment area of service, to cover intensive phase and weekly for those in the continuation
the three geographic areas of the project (lowlands, midlands, phase. During the second and third sessions, the “Other Visi-
and highlands). All three had existing access to a motorbike tors” registration mode (Figure 1) was used for demonstration
and were literate.‡ None of them had used a computer before. and individual practice. During the initial patient enrollment
at the household level, direct support of the research team was
immediately available to CHWs for questions related to patient
‡All CHWs in Uganda are literate. registration, following patient consent. The CHWs were allowed
1274 SNIDAL AND OTHERS

to proceed independently once they had demonstrated (Table 1), defined as cured, completed, transfer out, failure,
confidence with and complete knowledge of enrolling and death, or lost to follow-up.
observing patients, as determined by trainer evaluation Statistical methods and quantitative analysis. Data for the
(enrolling two patients without support). group enrolled on eCompliance was exported directly from
Subject selection and comparison cohorts. All adult patients the system and data for comparison groups was collected from
(> 18 years of age) enrolled into TB DOTS at clinics in the TB registers at the relevant clinics. All data was anonymized
MVP cluster during the pilot period (July 2012–December before being given to the research team for analysis.
2012) were offered enrollment to be followed by CHWs using The available summary statistics for the four cohorts were
eCompliance until the completion of their TB treatment reviewed to determine if there was significant imbalance among
(6 months duration). All gave verbal consent. No patient groups that might bias the interpretation of the results. The
refused to participate during or after enrollment. This cohort continuous variable, age, was compared across cohorts by using
of patients followed daily by CHWs using eCompliance will be 5-number summaries comparisons with T tests to assess dif-
referred to as MVP 2012. Impact was then evaluated by com- ferences of means. Categorical variables, gender, retreat-
paring MVP 2012 against patients undergoing TB DOTS in the ment status, type of TB, HIV status, if HIV counseling/testing
same rural district in Uganda (Isingiro). These non-intervention was received, if cotrimoxazole preventive therapy (CPT) was
comparison patients fell into three groups of patients: MVP started, and if antiretroviral therapy (ART) for HIV co-infected
2011, Isingiro 2011, and Isingiro 2012. MVP 2011 consisted of patients was started, were assessed and compared by running
TB patients enrolled on TB DOTS during 2011 (July 2011– crosstabs and Z tests for proportions.
December 2011) at the same cluster of clinics as the eCompliance Outcomes of observation were categorized into one or more
patients (i.e., MVP 2012). The latter two comparison cohorts possible results: cured, completed, treatment success, trans-
consisted of patients who received care outside of the MVP ferred out, failure, lost to follow-up, deceased, or negative
cluster of clinics, but in the same district as the MVP cluster outcome. Patients were then assigned a binomial result (yes or
clinics. Patients in the Isingiro cohorts lived in similar rural no) for each of these categories. The proportion of patients for
conditions as the MVP cohorts, with comparable barriers to cured, treatment success, lost to follow-up, deceased, and neg-
access as those in the Ruhiira cluster. Isingiro 2011 consisted ative outcomes§ were then calculated for all four groups. A Z
of patients enrolled into TB treatment during the time period test for proportion and case-cohort odds ratio (OR) compari-
of July 2011–December 2011 and Isingiro 2012 consisted of son was then carried out for each of those five outcomes in two
patients enrolled between July 2012–December 2012. All three comparisons: a test for statistical difference between MVP
comparison groups followed Uganda national TB guidelines 2011 and MVP 2012, and for a composite control group (“non-
and were supervised daily by family members, and once intervention”) and MVP 2012.
monthly by a CHW. All patients were followed until a treat- Finally, a difference in differences comparison was done to
ment outcome was available (Table 1) and were not blinded to assess the difference in change over time of outcome propor-
enrollment on the eCompliance intervention. tions in the MVP clusters as compared with the Isingiro clus-
Five data points of interest were collected for each patient: ters from the 2011–2012 cohorts. This was done by creating an
age; gender; if the patient was undergoing retreatment; HIV interaction term between year of treatment and which clinic
status and therapy received and the outcome of treatment the patient was enrolled in (year*mvp), then running a logis-
tical regression with OR.
All analyses were conducted in Stata Statistics 12.0
Table 1
(StataCorp, College Station, TX).
Outcome classifications*
Cured A patient whose sputum smear or culture was
positive at the beginning of the treatment but RESULTS
who was smear- or culture-negative in the
last month of treatment and on at least one Software adaptations. The eCompliance software was orig-
previous occasion. inally developed by Operation ASHA for use in urban set-
Completed A patient who has completed treatment but does tings in India. The implementation of the biometric system in
not have a negative sputum smear or culture
the new context resulted in changes to the coding of the
result in the last month of treatment and on
at least one previous occasion. software to follow Ugandan National TB Program guidelines.
Treatment success The sum of cure and completed. Four major updates to the software were made in the early
Transfer out A patient who has been transferred to another stages of project implementation (before enrollment of
recording and reporting unit and whose patients mid-July 2012). These updates addressed errors
treatment outcome is unknown.
Failure A patient who has a sputum smear or culture found in the coding during CHW training and early project
that is positive at 5 months or later during implementation and also provided updates to the system for
treatment. Also included in this definition are the change in cultural context. The first update fixed an
patients found to harbor a multidrug-resistant error that had caused patients to disappear after a status
(MDR) strain at any point of time during the
treatment, whether they are smear-negative
or -positive.
Death A patient who dies for any reason during the §The “cured” and “treatment success” outcomes were used as the
course of treatment. primary proportions for evaluation, “completed” is the difference of
Lost to follow-up A patient whose treatment was interrupted for the two and provides less information about project improvement,
2 consecutive months or more. therefore it was not separately assessed. The proportions for patients
Negative outcome The sum of failure, death, and default. transferred out were not analyzed because of the inability to know
*Definitions adapted from the World Health Organization (WHO) Guidelines for tuber- their official treatment outcome. Finally, failure was not evaluated
culosis treatment. individually as only one patient had this outcome in the cohort.
 USE OF TB eCOMPLIANCE IN UGANDA 1275

Table 2
Baseline characteristics and outcomes of the cumulative group and by cohort (N = 142)
n (%)

Cumulative control Isingiro 2011 Isingiro 2012 MVP 2011 MVP 2012
Characteristics All (N = 142) (N = 106) (N = 40) (N = 32) (N = 34) (N = 36)*

Sex
Male 104 (73.2) 79 (74.5) 30 (75.0) 23 (71.9) 26 (76.5) 25 (69.4)
Female 36 (25.4) 27 (25.5) 10 (25.0) 9 (28.1) 8 (23.5) 9 (25.0)
Unknown 2 (1.4) 0 (0) 0 (0) 0 (0) 0 (0) 2 (5.6)
Age
25% 30 30 31 30 31 28
Median 40 40 39.5 38.5 41 40
75% 49 48 49 45.5 46 52
Patient category
New 122 (85.9) 95 (89.6) 35 (87.5) 28 (87.5) 32 (94.1) 27 (75)
Retreatment† 18 (12.7) 10 (9.4) 4 (10) 4 (12.5) 2 (5.9) 8 (22.2)
Unknown 2 (1.4) 1 (0.9) 1 (2.5) 0 (0) 0 (0) 1 (2.8)
Type of tuberculosis
Pulmonary, sputum positive 113 (79.6) 87 (82.1) 36 (90.0) 27 (84.4) 24 (70.6) 26 (82.2)
Pulmonary, sputum negative 19 (13.4) 13 (12.3) 2 (5.0) 1 (3.1) 10 (29.4) 6 (16.7)
Extra-pulmonary 8 (4.9) 6 (5.7) 2 (5.0) 4 (12.5) 0 (0) 2 (5.6)
Unknown 2 (2.1) 0 (0) 0 (0) 0 (0) 0 (0) 2 (5.6)
HIV status
HIV-positive 55 (38.7) 43 (40.6) 15 (37.5) 12 (37.5) 16 (47.0) 12 (33.3)
HIV-negative 80 (56.3) 60 (56.6) 24 (60.0) 20 (62.5) 16 (47.0) 22 (61.1)
Unknown 8 (4.9) 3 (2.8) 1 (2.5) 0 (0) 2 (5.9) 2 (5.6)
Patient knows HIV status and received counseling
Yes 132 (93.0) 100 (94.3) 36 (90.0) 32 (100) 32 (94.1) 34 (94.4)
No 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0)
Unknown 10 (7.0) 6 (5.7) 4 (10.0) 0 (0) 2 (5.9) 2 (5.6)
Enrolled on CPT, where applicable (% total/percentage of applicable)
Yes 52 (36.6/94.5) 40 (37.7/93.0) 15 (37.5/100) 11 (34.4/91.7) 14 (41.2/87.5) 12 (33.3/100)
No 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0)
Unknown 3 (2.1/5.5) 3 (2.8/7.0) 0 (0) 1 (3.1/8.3) 2 (5.9/12.5) 0 (0)
Enrolled on ART, where applicable (% total/percentage of applicable)
Yes 21(14.8/38.2) 17 (16.0/39.5) 8 (20.0/53.3) 5 (15.6/41.7) 4 (11.8/25.0) 4 (33.3/11.1)
No 2 (1.4/3.6) 1 (0.1/2.3) 0 (0) 0 (0) 1 (2.9/6.3) 1 (2.8/8.3)
Unknown 32 (22.5/40.0) 25 (23.5/58.1) 7 (17.5/46.7) 7 (21.9/58.3) 11 (32.4/68.8) 7 (19.4/5.8)
Outcome
Cured 48 (33.8) 30 (28.3) 13 (32.5) 6 (18.8) 11 (32.4) 20 (55.6)
Success 108 (76.1) 79 (74.5) 28 (70.0) 26 (81.3) 25 (73.5) 29 (80.6)
Transfer 16 (11.3) 9 (8.5) 5 (12.5) 3 (9.4) 1 (2.9) 7 (19.4)
Failure 1 (0.01) 1 (0.9) 0 (0) 1 (3.1) 0 (0) 0 (0)
Lost to follow-up 8 (5.6) 8 (7.6) 5 (12.5) 0 (0) 3 (8.8) 0 (0)
Died 9 (6.3) 9 (8.5) 2 (5) 2 (6.3) 5 (14.7) 0 (0)
Negative outcome 18 (12.7) 18 (17) 7 (17.5) 3 (9.4) 8 (23.5) 0 (0)
*Enrolled on eCompliance.
†Retreatment, a patient who is returning after default, or receiving retreatment after initial treatment failure, or who has relapsed; for outcome definitions see Table 1.
Isingiro 2011 represents those patients enrolled on manually recorded DOTS at clinics outside of the MVP cluster in 2011; Isingiro 2012 represents those patients enrolled on manually recorded
DOTS at clinics outside of the MVP cluster in 2012; MVP 2011 represents those patients enrolled on manually recorded DOTS at clinics inside the MVP cluster in 2011; MVP 2012 represents those
patients enrolled on DOTS at clinics in the MVP cluster in 2012 where eCompliance was used as the method of recording observations; and cumulative control represents the composite of Isingiro
2011, Isingiro 2012, and MVP 2011.

transition—e.g., from intensive phase to continuous. The sis showed no significant relationship between age and any
second update fixed a glitch that caused the system to lose DOTS outcome so the patients were not removed (see Table 2).
its ability to acknowledge fingerprint scans after successfully Additionally, it was determined by T test that there was not a
registering a patient until it was restarted. The final two statistically significant difference in age distribution between
updates during the pre-pilot phase fixed a coding error that the enrolled patients and the non-intervention group (see
caused the system to occasionally stop being responsive. Dur- Table 3). There was also no statistically significant difference
ing the final updates, two minor changes were also made to following a Z test in the proportion of men to women. In an
accommodate the specific local Ugandan context: allowing assessment of the number of patients receiving retreatment
tracking for treatment on Sundays, which was not an option for TB, MVP 2012 had a significantly higher proportion of
in the Indian version of the system; and changing the wording retreatment patients than the cumulative control group
of patient registration to read “Scan patient’s finger” rather (22.2% as compared with 9.4%; P < 0.05); however, during
than “Scan your finger” to avoid any confusion. later analysis, no statistically significant effect of being a
The system was fully functional upon enrollment of patients. retreatment patient on positive outcomes of TB DOTS treat-
Demographics. No patients declined enrollment for follow- ment was observed, only a positive correlation with negative
up by a CHW using the eCompliance system. The total num- outcomes. Similarly, an assessment of the proportion of
ber of patients in all cohorts was 142. A 5-number summary patients according to type of TB, HIV status, and enrollment
showed that there were three patients who fell outside the on CPT and/or ART treatment showed no statistical differ-
1.5 times interquartile range of ages. However, further analy- ence between MVP 2012 and the control groups.
1276 SNIDAL AND OTHERS

Table 3 Table 4
Results of demographic and outcome comparisons Estimated magnitude of intervention effect
MVP 2012* vs. MVP 2012* vs. MVP 2012* vs. cumulative control MVP 2012* vs. MVP 2011
cumulative control MVP 2011
Outcome Odds ratio P value OR 95% CI Odds ratio P value OR 95% CI
P value P value
Cured 3.17 0.003† 1.35, 7.46 2.61 0.050‡ 0.891, 7.49
Demographics Success 1.49 0.484 0.42, 5.34 1.46 0.461 0.52, 4.27
Sex† 0.552 0.509
*Enrolled on eCompliance.
Age 0.375 0.779 †Statistically significant at P < 0.01.
Retreatment‡ 0.046§ 0.051 ‡Statistically significant at P < 0.05.
Type of tuberculosis An estimation of the magnitude of effect of the intervention using a case-cohort compar-
ison. OR = odds ratio; CI = confidence interval.
Pulmonary, sputum positive 0.205 0.880
Pulmonary, sputum negative 0.503 0.204
Extra-pulmonary 0.981 0.163
Unknown 0.015§ 0.163 When MVP 2012 patients were evaluated against non-
HIV status intervention patients, the difference in proportions of patients
HIV-positive 0.442 0.241 cured was found to be statistically significantly greater at
HIV-negative 0.636 0.238 55.6% compared with 28.3%, respectively (P < 0.05) (see
Unknown 0.443 0.953
Patient knows HIV Table 3). A case-cohort OR shows that MVP 2012 patients
status and received had 3.17 higher odds (P < 0.05) of having DOTS resulting in a
counseling cured outcome than all other patients (see Table 4). Similarly,
Yes 0.981 0.435 MVP 2012 had a statistically significant higher proportion of
No − −
patients with a cured outcome than MVP 2011: 55.7% and
Unknown 0.981 0.953
Enrolled on CPT, 32.4%, respectively (P < 0.05) (see Table 3). A case-cohort
where applicable OR shows that MVP 2012 patients had 2.61 higher odds (P <
(of all patients/ 0.05) of having DOTS resulting in a cured outcome than MVP
of those applicable) 2011 patients (see Table 4). Correspondingly, the logistic
Yes 0.636/0.347 0.497/0.204
No − − regression shows positive correlation between enrollment on
Unknown 0.308/0.347 0.140/0.204 eCompliance and odds of having a cure outcome on the
Enrolled on ART, cohorts over time, with log odds of 1.13 (P < 0.05) and an OR
where applicable of 3.09 (P < 0.05) greater likelihood of receiving a DOTS
(of all patients/of
“cure” outcome when in the intervention group (see Table 5).
those applicable)
Yes 0.472/0.696 0.932/0.630 The MVP 2012 had a higher proportion of patients classified
No 0.420/0.326 0.967/0.832 as “treatment success” (80.6%) compared with MVP 2011
Unknown 0.607/0.990 0.217/0.570 (73.5%) and all non-intervention patients (74.5%). However,
Outcomes neither difference between MVP 2012 and the non-intervention
Cured 0.003¶ 0.050§
Success 0.464 0.484 groups was statistically significant (see Table 3). Similarly,
Lost to follow-up 0.090 0.068 neither the logistic regression nor a case-cohort analysis
Dead 0.071 0.017§ shows a statistically significant correlation between being
Lost or dead 0.010¶ 0.002¶ enrolled on eCompliance and odds of having a DOTS “suc-
Negative outcomes 0.008¶ 0.002¶
cess” outcome (see Tables 4 and 5).
*Enrolled on eCompliance.
†Given records with unknown gender the values report reflect the most statistically signif-
Negative treatment outcomes. Negative outcomes appeared
icant Z and P values. in three forms in this population, as defined by the WHO
‡Retreatment, a patient who is returning after default, or receiving retreatment after
initial treatment failure, or who has relapsed. (Table 1): lost to follow-up, deceased, and treatment failure.
§Statistically significant at P < 0.05.
¶Statistically significant at P < 0.01. Only one patient in the group had an outcome of treatment
A comparison of proportions for the demographic characteristics and outcome indicators for failure, thus only the first two outcomes were compared on an
the MVP 2012 cohort, who received the treatment, against the proportions for the control
group in the MVP cluster in 2011 and for all of the patients in a control group, respectively. individual level. Additionally, because of perfect prediction
errors, no case-cohort or difference in difference analysis
could be calculated for the negative outcomes.
Of the 142 patients in the study, 36 (25.4%) were in MVP When MVP 2012 patients were evaluated against non-
2012, and therefore enrolled on the follow-up by a CHW intervention patients for the proportion lost to follow-up, the
using the eCompliance biometric system. Of all patients, 0% lost to follow-up rate in MVP 2012 was trending on signif-
108 (76.1%) had a successful treatment outcome (of which icantly lower than the 7.6% lost to follow-up rate of all other
50 patients, or 35.2%, were considered cured of TB), and
16 (11.3%) transferred to different clinics for treatment.
Of the remaining patients, 18 (12.7%) had negative outcomes Table 5
of treatment as follows: 9 (6.3%) died, 1 was deemed a treat- Logistic regression outcome using year*mvp
ment failure, and 8 (5.6%) were lost to follow-up (for out- year*mvp

come according to cohort see Table 2). Outcome Coefficient Odds Ratio OR P value OR 95% CI
Cured and treatment success outcomes. Positive outcomes Cured 1.13 3.09 0.006† 1.38, 6.9
appeared in two forms in this population, as defined by the Treatment success 0.56 1.76 0.293 0.61, 5.01
WHO (Table 1): “treatment completed” and “cured,” and the For outcome definitions see Table 1. The interaction term year*mvp compares the differ-
ence in change in proportion of a TB DOTS outcome from 2011 to 2012 between patients
aggregate of these two outcomes, defined as treatment suc- enrolled on eCompliance, and those not enrolled on the system. OR = odds ratio; CI =
cess. Evaluation was done only for cure outcomes and treat- confidence interval.
*Statistically significant P < 0.05.
ment success. †Statistically significant at P < 0.01.
 USE OF TB eCOMPLIANCE IN UGANDA 1277

cohorts (see Table 3). Similarly, MVP 2012 had a lost to to treat and usually have worse outcomes than new cases; yet
follow-up rate of 0%, which was trending on statistically dif- that cohort still had significantly better outcomes.
ferent when compared with the 8.8% lost to follow-up rate in The improvement in both the proportions of patients cured
MVP 2011 (see Table 3). and deceased were consistently the most statistically signifi-
When MVP 2012 patients were evaluated against the non- cant. Compared with all groups, MVP 2012 improved on both
intervention patients for proportion of death, MVP 2012 had of these indicators with statistical significance (P < 0.05), with
a lower proportion of death at 0% as compared with a 8.5% the exception of the death rate comparison with the non-
death rate for all other cohorts, trending on significant (see intervention group, which is only trending on significance.
Table 3). More strongly, MVP 2012s proportion of death was “Died” and “Cured” indicators have the most exactly defined
lower than MVP 2011s and statistically significant with 0% of measurements, and are the most difficult to improperly
patients dying and 14.7% in MVP 2011 dying (see Table 3). report. These results suggest a great potential for correlation
Overall, the reduction in negative outcomes was statistically between implementing a comprehensive DOTS program
significant on both comparisons as well. None (0%) of MVP using eCompliance and improving positive outcomes and
2012 patients logged a negative treatment outcome, whereas decreasing mortality. Our study was limited in size and these
non-intervention patients had a significantly higher negative outcomes could not be shown unequivocally. There is there-
outcome proportion of 17% (P < 0.05; see Table 3). Similarly, fore the necessity to conduct further larger studies using
MVP 2011 patients had a significantly higher negative outcome eCompliance to confirm these trends.
proportion of 23.5%, much higher than the 0% proportion of In contrast, the increase in proportion of patients with a
MVP 2012 (P < 0.05; see Table 3). treatment success outcome was not statistically significant.
Because “treatment success” combines patients who are
cured and who completed treatment, we believe this can be
DISCUSSION explained largely to a significantly higher proportion of com-
pleted cases in non-intervention compared with MVP 2012.
Health care workers monitoring DOTS, especially in clinics Those patients adhered to the full course of drugs, but did not
in resource-constrained settings, are rarely focused solely on get their final sputum checked for bacteriological confirma-
TB activities. Their responsibility to follow-up on the appro- tion of cure. Although the WHO target for treatment moni-
priate outcome of treatment is combined with a multitude of tors treatment success, the preferred outcome remains a
other tasks and obligations in other fields, such as malaria, demonstrated cure over a completed treatment, which was
HIV/AIDS and antenatal care, that are as important or time- statistically significantly higher in MVP 2012. It is also inter-
consuming. Although the “directly-observed” component of esting to note that, based on WHO recommendations, the
DOTS seems logical to prevent non-adherence to treatment, calculation of overall outcome of treatment still takes into
which may lead to increased risks of deaths and/or drug resis- account patients who have transferred out, even though these
tance for TB patients, it also places a substantial burden on patients are no longer under the care of the health system that
already overworked staff. As a result, patient TB cards sum- diagnosed them. If the calculation of treatment success were
marizing individual monitoring are inadequately completed, if based only on patients who completed treatment where it was
at all, and there is no simple system in place to help healthcare initiated, the treatment success for MVP 2012 would be 100%
workers effectively monitor TB patients under their supervi- (Table 2).
sion and prevent patients from being lost to follow-up. These results strongly suggest that there are improvements
We have piloted a home-based DOTS system employing on actual patient outcomes. Some patient outcome improve-
CHWs using a biometric system, originally designed for urban ments are expected because patients who were adequately
slums in India, and introduced it in a rural village in Uganda. monitored on a daily basis, with the support of CHWs and
This system, entitled eCompliance, facilitates the monitoring eCompliance, did not skip pills, did not default, and therefore
of TB patients adherence to treatment by using fingerprints at had a higher chance of completing treatment. Larger follow-
a netbook terminal during every visit, whether in the home or up studies would be very valuable to confirm these indica-
at a clinic. The system was easily adapted to Ugandan National tions, particularly the correlation between the use of the
TB Program requirements of daily supervision of DOTS eCompliance system and the decrease in patient mortality.
through the full course of TB drugs, and was fully functional Additionally, eCompliance provides the opportunity for
within a couple weeks, before enrollment of patients. the CHWs and any monitoring team to collect more detailed
Compared with non-intervention patients, the patients and verified dose observation records for each of the patients
monitored by CHWs using eCompliance (MVP 2012) had enrolled on eCompliance: total number of missed doses; the
improved treatment outcomes at multiple levels. The differ- number of individuals with missed doses; the number of indi-
ence in the rates of negative outcomes for MVP 2012 (0%) viduals with repeated missed doses; and the number of indi-
was significantly lower than that for the non-intervention viduals that had missed 2 or more consecutive weeks of doses
cohorts (17%) and when compared just to those patients (and therefore were potentially believed to be at risk of
treated at clinics in the MVP cluster (23.5%). Similarly, the discontinuing TB treatment). This information provides more
rate of patients cured of TB following DOTS was signifi- insight into how regularly patients are missing doses and how
cantly higher for MVP 2012 (55.6%) when compared with often consecutive missed doses lead to being at risk for non-
non-intervention cohorts (28.3%), including when compared adherence. This information is particularly useful to health
with only patients treated at the same cluster of clinics (32.4%) workers as it allows them to easily assess which patients require
the previous year. It is also important to note that MVP 2012 extra support and education during care. We believe that the
had a proportion of retreatment patients significantly higher large gains made on patient outcomes during the pilot are
than other groups; these patients are typically more difficult largely a result of the support that the eCompliance system
1278 SNIDAL AND OTHERS

gives to health care providers to ensure they are reaching every wherever they go, or if an absent CHW needs to be replaced by
single patient for TB care and support. In this case, we believe a colleague using a different machine. We expect that these
that better, more accurate documentation that is also easily and options would further support efficiency of TB DOTS pro-
quickly accessible by health teams can ultimately lead to grams, particularly as they allow for adaptation of eCompliance
improved patient support and, therefore, better outcomes. within various health systems, and being viable options in rural
Although available, we did not present these details for our areas because they communicate through an SMS-based plat-
cohort in this study because they were difficult or impossible form, and do not required internet access.
to retrieve for the patients in non-intervention cohorts. This Similarly, this pilot project limited the use of eCompliance
suggests the importance of a system such as eCompliance that and the biometric medical record system to TB patients.
seamlessly tracks all these additional indicators for each However, given the success, we expect that the eCompliance
patient. The system could also be updated to collect further technology, using modified software, could provide support to
information for patient records, such as HIV status and, if other challenges in health programs in resource-limited set-
applicable, enrollment on ART and CPT. This would allow tings. These functions could be as straightforward as monitor-
CHWs to have a more complete view of each patient, and ing the follow-up of other health programs, such as antenatal
quickly assess if patients are at higher risk of being lost to care, antiretroviral treatment, and primary care. Additionally,
follow-up. Additionally, it could allow for TB and HIV pro- the instantaneous and accurate information regarding patient
grams to be further linked to support co-infected patients. loads and TB burdens can be used to predict reagent, commod-
It may be interesting in further studies using eCompliance ity, and drug needs and direct resources to where they are
to add a qualitative assessment regarding why patients miss needed most. Streamlining information about supply chain
doses, particularly when they miss consecutive observations. needs ensures that all patients are provided medicines improv-
This information would provide further insight into potential ing treatment completion. Additionally, the system provides
health system improvements, especially if there is a pattern additional security for patient information because it requires
regarding when patients would require additional education biometric login of health care workers to access patient infor-
and/or support during their time on DOTS. mation and history. Once stored in the system, the biometric
Though there was no formal qualitative survey of the information is securely encrypted, and cannot be exported out
eCompliance pilot project, feedback from the CHWs and of the eCompliance system, preventing fraudulent use.
patients was largely positive. Both the CHWs and the patients Limitations. The sample size used for this study was rela-
commented that the system was helping in fully following up tively small. Although still considered large enough to show
with patients, and therefore were more likely to have positive statistical significance, a larger sample size would have pro-
treatment outcomes of TB DOTS. The CHWs felt that vided more insight into the effect of enrolling patients on
eCompliance helped them to be more efficient and to increase eCompliance.
community and patient confidence in the health program. Our intervention revolved around daily supervision by a
Patients reported that the system had reassured them that they CHW using eCompliance, whereas the control groups relied
were being properly cared for and that their TB treatment on daily supervision by a family member, with a monthly visit
would be correctly monitored in its entirety. Additionally, no by an appointed CHW. We chose this protocol for the inter-
patients or CHWs expressed concern about added stigma vention cohort because daily supervision represents the most
related to use of eCompliance, even when prompted. Con- stringent of conditions for implementation of DOTS, and in
versely, in the clinics outside of the MVP cluster, where several countries, daily supervision is the recommended treat-
eCompliance was not implemented, the CHWs were frustrated ment protocol. We showed that this frequency is achievable
with the current inefficacy of the DOTS program in adequately and does not constitute a burden for the health workers.
monitoring patients—and even felt they may have been doing However, the improved outcomes of the intervention group
the patients a disservice. Future reviews of eCompliance may could be attributed to the increased frequency of a visit by a
wish to formally review CHW and patient perceptions of the CHW. Because it was already shown that direct observation
system, and changes in satisfaction of the associated health- for DOTS can be successfully undertaken by either a family
care system. member or a CHW,11,12 a larger follow-up study to our pilot
The effectiveness of eCompliance in this pilot project should segregate daily supervision by a CHW from daily
focused solely on the use of patient outcomes as indicators. supervision by a CHW using eCompliance to determine the
However, it may also be interesting to evaluate eCompliance added impact of the biometric monitoring tool.
based on its potential increase in health worker efficiency. Additionally, the quality and quantity of control group data
Because health workers are able to easily access patient sched- points was fairly limited for analysis. Although eCompliance
ules, view a verified record of observations, and quickly log collects and verifies each dose and outcome through its bio-
accurate patient observations, eCompliance may reduce the metric technology, the control group data is subject to human
amount of time spent on paperwork, facilitating the timely error. The data was also only outcome based and did not look
follow-up of patients. In addition, eCompliance can also send at the progression of treatment or the individual patient level
daily text message to CHWs informing which patients require data because of this limitation. Similarly, patient level data
follow-up, though that feature was not used during this pilot. It such as missed doses was available for those enrolled on
also allows for the option that multiple systems can synchro- eCompliance, and yet could not be used as a comparison
nize, therefore a patient could potentially be enrolled on mul- because information was limited from the control groups. The
tiple machines and transfer between clinics or CHWs while nature of the population, in terms of which variables could be
maintaining complete and accurate DOTS records. This fea- controlled for, was limited as well: for instance, we did not have
ture would be particularly interesting for patients visiting more access to factors such as education level, nutritional intake,
than one local clinic for care as their profile would be available household income, etc., which may have influenced outcomes
 USE OF TB eCOMPLIANCE IN UGANDA 1279

at the patient level. However, by comparing the eCompliance would support the goal of reducing drug resistance and mini-
data against outcomes for patients taken from the same popu- mizing TB mortality.
lation within the Ruhiira cluster, and from neighboring com-
munities in Isingiro district, we believe that these factors Received July 2, 2014. Accepted for publication January 31, 2015.
should be fairly controlled. Additionally, when considering the Published online April 6, 2015.
success of the eCompliance project, the quality of CHWs must
Acknowledgments: We are grateful to Robert Mugabe, John Asiimwe
be considered. Across the Ruhiira cluster cohorts, this should Museveni, and Sam Twinomujuni for their help with the implementa-
be standardized as they all receive the same training and are tion of the eCompliance system. We also thank Anne Liu, Saira
held to the same accountability. Finally, this study did not take Qureshi, and Jilian Sacks for critical reading of the manuscript.
into account the clustering effect on CHWs, which should be Financial support: This publication was supported by the National
considered in a larger follow-up study. Center for Advancing Translational Sciences, National Institutes of
Also of note, the high rate of patients that transfer to new Health (NIH), through Grant no. UL1 TR000040. The content is
clinics provides a challenge for all DOTS programs, and is a solely the responsibility of the authors and does not necessarily rep-
resent the official views of the NIH.
limitation to understanding the actual outcomes of patients
observed in this study. However, given the ability to transfer Disclaimer: All authors declare that they have no financial or non-
financial competing interest.
patient information and synchronize systems using the bio-
metric technology, this would not be a challenge should Authors’ addresses: Sarah Jane Snidal, LifeNet International, Global
eCompliance be used at a national level as clinics can easily Health Corps, Bujumbura, Burundi, E-mail: sarahjane.snidal@
gmail.com. Genevieve Barnard, The Schlesinger Fund for Global
connect to verify patient transfer and adherence. eCompliance Healthcare Entrepreneurship, Babson College, Wellesley, E-mail:
was designed to identify patients locally on the machine, which [email protected]. Emmanuel Atuhairwe, Millennium Villages
makes it a distributed system. A distributed system consists of a Project, Mbarara, Uganda, E-mail: [email protected].
collection of autonomous computers, connected through a net- Yanis Ben Amor, The Earth Institute, Columbia University, NY,
E-mail: [email protected].
work, which enables computers to coordinate their activities
and to share the resources of the system, so that users perceive
the system as a single, integrated computing facility. Addition- REFERENCES
ally, each machine supports up to 10,000 fingerprints with iden-
1. WHO, 2012. Global Tuberculosis Report. Geneva, Switzerland:
tification speed of 3,000 templates per seconds at more than
WHO. Available at: http://www.who.int/tb/publications/global_
99.5% accuracy. eCompliance can therefore be scaled up to a report/en/. Accessed June 25, 2014.
national level. 2. Zumla A, Raviglione M, Hafner R, von Reyn CF, 2013. Tubercu-
losis. N Engl J Med 368: 745–755.
3. Gray JM, Cohn DL, 2013. Tuberculosis and HIV coinfection.
CONCLUSIONS Semin Respir Crit Care Med 34: 32–43.
4. Lonnroth K, Jaramillo E, Williams BG, Dye C, Raviglione M,
The TB eCompliance pilot project at the Ruhiira site shows 2009. Drivers of tuberculosis epidemics: the role of risk factors
and social determinants. Soc Sci Med 68: 2240–2246.
that there is transferability of the system to rural and 5. Zumla A, Nahid P, Cole ST, 2013. Advances in the development
resource-limited settings for use by community health of new tuberculosis drugs and treatment regimens. Nat Rev
workers, with adaptations of the software to local context. Drug Discov 12: 388–404.
Indeed, the various DOTS projects across continents will 6. Glaziou P, Falzon D, Floyd K, Raviglione M, 2013. Global epide-
require that the eCompliance system be properly adjusted to miology of tuberculosis. Semin Respir Crit Care Med 34: 3–16.
7. Ben Amor Y, Day MS, Schluger NW, 2010. Preventing the next
fit within the individual country setting. However, this can be generation of extensively drug-resistant tuberculosis. Int J
achieved with basic software updates as was accomplished for Tuberc Lung Dis 14: 525–527.
the pilot project in Uganda, ensuring that the specifications of 8. Lienhardt C, Glaziou P, Uplekar M, Lonnroth K, Getahun H,
the system properly match the local and national TB pro- Raviglione M, 2012. Global tuberculosis control: lessons learnt
and future prospects. Nat Rev Microbiol 10: 407–416.
grams. It also shows that the system can be used successfully 9. Davies PD, 2003. The role of DOTS in tuberculosis treatment
in a non-clinic based model, and when managed by CHWs. and control. Am J Respir Med 2: 203–209.
Unique challenges, such as finding available charging stations 10. Operation ASHA, 2013. Technology: The eCompliance Project.
and transport to patients’ homes, exist. However, where func- Available at: http://www.opasha.org/our-work/ecompliance-
tioning health programs exist, the eCompliance system can innovation-and-health/. Accessed October 11, 2013.
11. Newell JN, Baral SC, Pande SB, Bam DS, Malla P, 2006. Family-
overcome these challenges. member DOTS and community DOTS for tuberculosis con-
Ultimately, the eCompliance pilot project showed that this trol in Nepal: cluster-randomized controlled trial. Lancet 367:
system can help improve follow-up of TB patients in a 903–909.
resource-limited rural setting, thereby increasing positive TB 12. Wright J, Walley J, Philip A, Pushpananthan S, Dlamini E,
Newell J, Dlamini S, 2004. Direct observation of treatment for
DOTS outcomes and decreasing negative outcomes. By tuberculosis: a randomized controlled trial of community
implementing eCompliance in resource-limited settings, the health workers versus family members. Trop Med Int Health
occurrence of patients lost to follow-up may be halted, which 9: 559–565.