Health IT
testing translation
Introduction
Tuberculosis (TB) is a chronic infectious disease that kills over two million people per year in the developing world. TB can typically be diagnosed rapidly by sputum microscopy at a local health center (HC), but diagnosis of multi-drug resistant TB (MDR-TB) – defined as TB strains resistant to at least isoniazid and rifampicin – requires a drug susceptibility test (DST) which is usually performed at a district, national or even supranational level. The emergence of extensively drug-resistant tuberculosis (XDR-TB) heightens the urgency of prompt diagnosis of drug resistance to curb the excessive mortality and ongoing transmission associated with highly resistant strains.1 Prompt treatment with individualized drug regimens based on DST improves patient outcomes2 and reduces the risk of amplification of drug resistance and ongoing transmission.3 4
To improve detection and treatment of MDR-TB and XDR-TB, the Peruvian Ministry of Health has decentralized rapid and conventional 1st line DSTs from the National Reference Laboratory (NRL) to several district laboratories.5 However, like many other countries, communication of these DST results between central and local laboratories and clinical facilities is still problematic. A baseline assessment found that 10% of results take over two months to arrive and patients could still experience risky delays unless programmatic aspects are also addressed.6 To reduce these delays, we developed and implemented a laboratory information system, e-Chasqui, to communicate data between the NRL, two district laboratories and 12 HCs in Lima, Peru.7 This system has been shown to decrease the number of reporting errors to HCs by up to 87%, most importantly eliminating missing results.8
Laboratory information systems in developed countries have been shown to decrease turn-around-times (TAT) of laboratory results,9-11 reduce redundancy in resource utilization,10 12 13 and provide faster and more complete notification for public health purposes, though no randomized trial has shown a clinical impact.14-16 Shorter TATs have been associated with decreased treatment time, mortality, morbidity, and length of hospital stay.17 18 However, a systematic review found no reports of evaluations of these systems in resource-poor settings.19
We conducted a cluster randomized controlled trial to evaluate the effectiveness of the e-Chasqui laboratory information system in reducing communication delays, time to treatment, and time to the end of disease transmission (culture conversion) within the TB program in Peru.
Methods
A cluster randomized controlled trial (RCT) tested the effect of the laboratory information system e-Chasqui in reducing the time to communicate patients’ test results, start them on appropriate treatment, and achieve culture conversion. The trial is reported according to the CONSORT statement.20 This trial was performed within a larger observational study evaluating the impact of expanded laboratory capacity in the district laboratories.6 All data were collected prospectively. This study was approved by the Partners Healthcare Human Research Committee and the Peruvian National Institute of Health, and has been registered in ClinicalTrials.gov with identifier NCT01201941.
Study Settings
This study was carried out in two health districts of Lima, Peru: Lima Ciudad and Lima Este. Lima Ciudad includes 45 health establishments (24 HCs, nine health posts, and 12 hospitals) serving a population of 1,577,090 in an area of approximately 100 km2. Lima Este includes 134 health establishments (42 HCs, 87 health posts, and 5 hospitals) serving a population of 1,088,515 in an area of approximately 6340 km2. Smear microscopy is used to diagnose active TB, while culture and drug susceptibility testing (DST) are reserved for individuals with confirmed TB and at least one risk factor for MDR-TB according to National Tuberculosis Program (NTP) Norms.21 Smear microscopy is performed in Level I laboratories in HCs and hospitals. Health posts send sputum samples to their closest HC for smear microscopy. For patients with MDR-TB risk factors, smear-positive samples are sent to district laboratories for culture and/or DST to first-line drugs. Strains resistant to isoniazid or rifampicin or both are sent to the National Reference Laboratory (NRL) for DST to second-line drugs. Paper results are emitted from the NRL and district laboratories and transmitted, directly or indirectly, to health establishments. The patient is then routinely seen by a pulmonologist at the local hospital to review the DST results and, if necessary, modify the TB regimen. In patients with drug-resistant isolates, a district expert committee must review the case and approve initiation of MDR-TB therapy (Figure 1).
The two health districts organize transmission of paper results to HCs differently. In Lima Ciudad, all 24 HCs are point of care HCs that receive results directly from the district laboratory. In Lima Este, health establishments are organized in “micro-networks.” Seventeen point of care HCs serve as the micro-network heads, and use the identical process as HCs of Lima Ciudad. The other 25 HCs and 87 health posts are peripheral HCs, which receive test results via the head of its micro-network (Figure 1).
Study Design
In March, 2006, e-Chasqui was first implemented at the two district laboratories and the NRL. These laboratories served all of the health establishments. After full implementation in the laboratories, 12 of 34 point of care HCs were randomized to utilize e-Chasqui. Health centers were randomly assigned following simple randomization procedures to intervention or control groups, with no allocation concealment. The allocation sequence was generated by two investigators and assigned to each health center. In Lima Ciudad, we randomly assigned e-Chasqui to six of the 20 highest TB incidence HCs in Lima Ciudad and to six of the 12 Lima Este micro-networks within Lima city limits (Figure 1). In Lima Este, the six micro-networks assigned to e-Chasqui consisted of six point of care HCs with 17 peripheral HCs. The six control micro-networks were comprised of 6 point of care HCs with 27 peripheral HCs.
We performed intent-to-treat analysis based on the original HC randomization. During the study (October 2006), the Lima Este health district re-organized their micro-networks which had the following impact on study assignments: one point of care control HC became a peripheral intervention HC; three peripheral intervention HCs became peripheral control HCs; and three peripheral control HCs became peripheral intervention HCs. Data from these “cross-over” HCs only affected 17 data points for the primary outcome.
Study Population
All individuals who lived within the catchment area of participating health centers – the 20 HCs with highest TB incidence in Lima Ciudad and establishments within the 12 micro-networks within Lima Este city limits – and had at least one MDR-TB risk factor as defined by the Peruvian NTP Norms were included in this study.21 There were no exclusion criteria for enrollment into the study. Since sputum samples of patients in the public health sector with least one MDR-TB risk factor should all be submitted to the district laboratory for DST, subjects eligible for enrollment into the study were identified using only by this referral.
Outcomes
The primary outcome of the study was the laboratory turn-around-time (TAT), defined as the number of days between a test result date and the date that result was received by the HC (Table 1). For the electronic system, the date received at the HC was the earliest of the reception of the paper result as shown in the reception stamp by the HC or when the result was viewed online by a TB staff member. For the paper system, the date received at the HC was the reception of the paper result as shown in the reception stamp by the HC. This primary outcome was calculated for both cultures and DSTs. Secondary outcomes can be seen in Table 1.
Intervention
We designed and implemented a web-based laboratory information system “e-Chasqui” in Lima, Peru to improve the timeliness and quality of laboratory data.7 It was deployed in the NRL, two district laboratories, and 12 intervention HCs. The core of the e-Chasqui interface is a single patient page containing the history of all tests performed for the patient on a left sidebar, and the details for any single sample on the main part of the page. Tools built for the laboratory include quality control, reports on tests performed, warnings for delayed reporting of results, and a user directory to control any individual’s access. Tools for clinicians included email notification of new results, a consolidated list of results for their jurisdiction, and a list to track the status of all pending samples. All intervention HC staff were trained at their HC in an initial session for approximately one hour. The data administrator would then visit or call the HC at least twice a month and could be contacted via cell phone or email during business hours.
e-Chasqui was built as a stand-alone module on the open source Partners In Health Electronic Medical Record (PIH-EMR), a web-based system designed for TB and MDR-TB treatment in resource-poor settings.22 The PIH-EMR provides the ability to register patients, order medications23, display chest x-rays, generate monthly reports for funders, and predict future drug requirements.24 This system is built using Java, HTML and an Oracle® database, and has been adopted as the official system of the Peruvian NTP.
Sample Size
Previously we measured the treatment TAT to be approximately 65 days.6 Assuming that the effect estimate of e-Chasqui would reduce this delay by 20 days, based on 0.8 power, and an of 0.05, 165 subjects in each group (330 total) were required.
Data Abstraction
Baseline data were collected 15 months prior to the implementation of e-Chasqui (Jan 1, 2005-Mar. 30, 2006 for Lima Ciudad, May 1, 2005-Aug. 18, 2006 for Lima Este). However, the Lima Este district laboratory did not perform DST before the implementation of e-Chasqui, hence there are no pre-implementation data on DSTs for that district.
Data were prospectively abstracted by a team of trained collectors who used standardized forms. For the RCT, the study started on the date of implementation of e-Chasqui and ended on August 31, 2008. We also used data from the e-Chasqui database, including the date of electronic receipt of test result. If the end date for any TAT was missing, we censored that time using the date the patient finished the study.
Statistical Analysis
We examined the effect of the intervention at a sample and an individual level, adjusting for the impact on variance of the clustering in the study design. We used multivariate regression models (marginal model with generalized estimating equations) to investigate the effect of the intervention on the TAT outcomes as a function of covariates and to account for the clustering at the HC level.25 To investigate whether the intervention was associated with a reduction in the number of DST results with laboratory TAT greater than 60 days, we used a generalized linear mixed model26 27 with HC as a random effect and health district and period (pre- and post-implementation) as fixed effects.
To adjust for possible HC differences that may have been unequally distributed despite randomization, we included the median pre-intervention TAT per HC for each of the TAT outcomes (as a proxy for HC variance) and number of HC staff changes. At the individual level for the treatment and culture conversion TAT, we also adjusted for HIV and pediatric status. We used SAS version 9.1 (SAS Institute, Cary, NC, USA) for all analysis and checked all models built using R.28
Results
During the trial, 89% (1671/1888) of all eligible patients were enrolled (Figure 2). The intervention HCs had a significantly greater number of study participants per HC. If separated by district, these differences existed only in Lima Ciudad. The intervention HCs also had younger participants, more female participants, and a larger number of personnel changes in the TB clinician per HC. There were no significant differences in the number of patients per peripheral HCs; number co-infected with HIV; turnover of TB nurses during the study; number of patients with baseline positive smear or culture status or with drug-resistant TB by study arm (Table 2). 98% of all culture results and 100% of all DST results available in e-Chasqui were viewed by the intervention HCs.
TAT Outcomes
Intervention HCs took significantly less time to receive both DST (median 11 vs. 17 days, p<0.001) and culture (5 vs. 8 days, p<0.001) results (Table 3). Intervention HCs had 47% fewer DSTs with a laboratory TAT of greater than 60 days compared to control HCs, however this was not significant (p=0.12). For a total of 252 participants (134 in the intervention HCs, 122 in the control) treatment TAT did not significantly differ in the intervention versus control HCs (median 77 v. 88 days, p=0.28).
Among 247 participants included in analysis for culture conversion TAT (126 in intervention HCs, 121 in control), those in the intervention HCs had 20% lower time to culture conversion than those in the control HCs (p=0.047, Table 3).
Discussion
The e-Chasqui laboratory information system considerably reduced the time to communicate results of cultures and DST to local HCs and the proportion of results that had an excessive delay or never arrived, though this last value was not statistically significant. Although intervention versus control groups did not significantly differ in time to treatment, those in the e-Chasqui group did have a significant decrease in the time to culture conversion, compared with the control group.
The prospective, randomized nature of this trial allowed for rigorous evaluation of the effect of e-Chasqui within the National TB Program. There have been no prior evaluations reported of the impact of an electronic system in decreasing delays in a resource-poor clinical setting and the results of this study show that it can have a large effect in the communication time of critical laboratory data. This effect might be even greater if this system were used in rural areas since the obstacles to communication usually encountered (long travel times, infrequent transport, or weather) are easily surpassed if there is a reliable internet connection. Further, this system can prevent patients from “falling through the cracks.” Because laboratory networks typically report the number of MDR DSTs and the TB program reports the total number of patients on treatment, without an integrated electronic system, it is difficult to identify how many confirmed TB patients are not on treatment. This “break” in the patient care process can be easily overlooked when evaluating a TB program.
Few prospective, randomized trials have shown that an information system can have a clinical impact. Here, we found that the patients in the intervention HCs achieved culture conversion 16 days earlier than those in the control HCs, a 20% decrease. The mechanism of this impact, however, is unclear since it was not due to an earlier start of appropriate treatment. We believe that the culture conversion TAT measures not only the effect of the drug regimen, but other factors that we did not measure directly. Some of the ways e-Chasqui could contribute to the clinical impact are: 1) significantly reduce the number of DST results that are never received at the health establishment;8 2) improved monitoring of patients because clinicians have greater access to their bacteriological history; 3) increased ability to prioritize regimen changes for the patients who would benefit most; and 4) improved adherence by patients because they believe they are receiving better treatment when their doctor uses the e-Chasqui.
Another possible measure of the success of this system is its expansion and adoption by the Peruvian Ministry of Health. Currenlty, the use of e-Chasqui has been expanded to serve a network of 259 institutions serving a catchment area of over 4.2 million people and providing treatment to approximately 9,600 TB and 1,100 MDR-TB patients every year. Since the end of the study, the laboratories have taken over the duties of data management and the Peruvian non-profit Socios en Salud Sucursal Peru provides technical support to maintain the system.
There were fundamental baseline differences between the intervention and control HCs despite the randomized nature of this trial. These differences could introduce bias into the analysis, but we used pre-implementation values in our models to account for measurable confounders. The study was conducted in the two most populous health districts in Peru. Therefore the generalizability of these results should be treated with caution. Being in an urban area provided the project with mostly consistent power and internet, as well as geographic proximity to provide technical support, which could be more limited in other settings. Therefore groups implementing these systems should ensure that the appropriate infra-structure is in place. Finally, this was a formative, rather than summative, evaluation since the developers were involved.
This system only directly intervened in the first step of a patient's treatment which was the communication of results to clinical staff. We believe that the effect of a system could be expanded by incorporating additional system tools, such computerized initiation of regimen changes and tools to aid follow up of patients. This shows the great potential that even limited electronic systems that support the public infrastructure have in improving healthcare. Further, similar tools are being built in open source systems such as OpenMRS (www.openmrs.org) which can be downloaded from the internet.
Conclusion
A carefully designed and implemented web-based tuberculosis laboratory information system reduced the time to communicate results between laboratories and health establishments spread throughout a large, peri-urban area. Patients in intervention and control HCs did not significantly differ in time to treatment, but time to culture conversion was 18 days faster, compared with those in control HCs (21% earlier). This system has been expanded to over 250 health centers serving a population of 4.1 million. Such a system in other resource-poor settings should be considered as a component of laboratory infrastructure to support TB and MDR-TB care.
Edit Comment Text