Menu
Menu
Menu
Menu
Background: Previous operational research studies have demonstrated the feasibility of large-scale public sector ART programs in resource-limited settings. However, organizational and structural determinants of quality of care have not been studied. Methods: We estimate multivariate regression models using data from 13 urban HIV treatment facilities in Zambia to assess the impact of structural determinants on health workers’ adherence to national guidelines for conducting laboratory tests such as CD4, hemoglobin and liver function and WHO staging during initial and follow-up visits as part of Zambian HIV care and treatment program. Results: CD4 tests were more routinely ordered during initial history and physical (IHP) than follow-up (FUP) visits (93.0 % vs. 85.5 %; p<0.01). More physical space, higher staff turnover and greater facility experience with ART was associated with greater odds of conducting tests. Higher staff experience decreased the odds of conducting CD4 tests in FUP (OR 0.93; p<0.05) and WHO staging in IHP visit (OR 0.90; p<0.05) but increased the odds of conducting hemoglobin test in IHP visit (OR 1.05; p<0.05). Higher staff burnout increased the odds of conducting CD4 test during FUP (OR 1.14; p<0.05) but decreased the odds of conducting hemoglobin test in IHP visit (0.77; p<0.05) and CD4 test in IHP visit (OR 0.78; p<0.05). Conclusion: Physical space plays an important role in ensuring high quality care in resource-limited setting. In the context of protocolized care, new staff members are likely to be more diligent in following the protocol verbatim rather than relying on memory and experience thereby improving adherence. Future studies should use prospective data to confirm the findings reported here. © 2012 Deo et al.; licensee BioMed Central Ltd.
Of the more than 150 HIV care and treatment facilities run by MOH in Lusaka Urban District, we focused on 13 because of the similarity in geographic location, conditions of service-delivery for these clinics and, availability of data on staff burnout in these facilities from a healthcare worker survey [9] conducted between March and June 2007. As most of these clinics had been providing HIV care for over 6 months (one clinic only 5 months), the effects of initial scale-up were minimized. Our study was approved by the institutional review boards at Northwestern University, University of Alabama at Birmingham, and the Research Ethics Committee of the University of Zambia. We included all recorded visits of adult patients (aged 16 years or more) diagnosed with HIV and enrolled in care during the calendar year 2007. This coincides with the time frame of the healthcare worker survey [9]. Data on predictors related to staff motivation (burnout, experience, absenteeism and turnover), were obtained from the healthcare worker survey [9] in various primary health departments (maternal and child health, outpatient, inpatient, labor, HIV care and treatment, tuberculosis) in Lusaka Urban District in 2007. Approximately 500, anonymous responses from 13 facilities were received and analyzed. Some of the relevant questions are reproduced in Table Table11. Description of predictor variables and their data sources In addition to obtaining data on the outcomes of the study, electronic medical records (SmartCare database) were used to obtain data on the number of daily visits to facilities. The study did not require identification of patients; rather the primary record marker was the type of visit – ART initiation, ART follow up etc. Visit information was extracted based on inclusion / exclusion criteria mentioned above and all patient identifiers were removed. Records of overtime payments were used to calculate the number of shifts worked by nurses and clinical officers in ART clinics. In 2007, all staff members working in the ART clinics were deputed from other departments and worked overtime in the ART clinics. Administrative databases within CIDRZ and MOH were used to calculate the length of time since initiation of ART program services at each facility. Architectural plans of each facility were used to calculate the total floor area of each clinic, to determine calculation of physical space. Owing to shortages of physicians, clinical officers (analogous to physician’s assistants in the U.S.) and nurses delivered majority of healthcare services in our setting. To ensure a minimum standard of care, clinicians followed visit-specific protocols (initial visit form, routine follow-up form etc.) that reflected national treatment guidelines. These forms were designed to guide clinicians through initial evaluation of newly enrolling patients and all subsequent follow-up visits. In this study, we constructed dichotomous variables to indicate whether appropriate laboratory / clinical tests were conducted at each of these visits. Initial History and Physical (IHP) visit: At the initial visit, we examined whether patients were correctly assessed according to WHO and national guidelines. In addition we examined whether baseline laboratory investigations, such as CD4, hemoglobin, liver function tests were properly carried out in accordance with WHO staging and national guidelines. We gave a positive credit to the facility if results were recorded in the patient’s chart within 4 weeks before and after the patient’s visit. Follow-up (FUP) visits: We chose CD4 test as a measure of adherence to follow-up visit protocol because CD4 count is a key clinical indicator of treatment response (for those on treatment) and disease progression (for those not on treatment). Thus, it provides a better measure of adherence to protocol across all patients compared to other non-compulsory tests such as hemoglobin and liver function [22]. We measured whether repeat CD4 testing was ordered within 6 months of the previous test. To account for variability in patient attendance to scheduled visits, we developed the following rule. For each follow-up visit, we expected a CD4 count test to be done if there was no CD4 result entered in the database in the preceding 160 days. For each visit where CD4 was expected, we considered the CD4 test done if the result was recorded in the patient’s chart within four weeks after the visit. Our quality measure was calculated as the total CD4 tests done in each month, divided by the sum of visits where a CD4 was expected and not expected but done in each month for each facility. For sensitivity analysis, we repeated the analysis with 180 days time window. We also repeated the analysis with a different definition of done as either a tick mark on the patient’s chart or a result within 4 weeks. In addition, we also counted the CD4 test as done in follow-up visits where it was not expected according to our definition above. We did not have access to the identity of health care workers involved in provision of care to individual patients. Thus, we could not analyze the difference in adherence to protocols at the worker level but could only infer these differences at the facility level. Facility level measures for staff burnout, staff experience, staff absenteeism and staff turnover were calculated by taking the median of individual responses to the staff motivation survey from that facility. Monthly staffing ratio was calculated by dividing the patient visits to the ART department by full time equivalent (FTE) staff shifts. This included nurses, physicians, clinical officers, technicians and pharmacists. A measure of physical space was calculated by dividing the floor area of ART department by the total patient visits during 2007. An alternative measure using average patient visits per day for each month did not alter the results substantially. In our setting, more than 90 % of the space was used for delivery of care and the rest for administrative tasks. Clinic age was calculated as the time since the initiation of ART program services as of January 2007 (Table (Table2).2). All predictors, except those constructed from health worker survey, pertained to ART services. The predictors derived from the survey included staff members belonging to other departments as well. Summary statistics of predictor variables Each column shows the median value for each variable by site. Measures of these predictor variables were held constant for each site during the study period. We ran multi-level logistic regression models using SAS GLIMMIX procedure for visit level outcome variables. We used facility-month combinations to define the hierarchical structure, intercept as a random effect, and other predictors as fixed effects. We developed and analyzed two model variants (nested within each other) to assess the incremental impact of different predictor variables: (i) Model 1 included Staff Experience, Staff Turnover, Space per visit, Clinic Age, Visits per shift, (ii) Model 2 included all the above predictors and Staff Absenteeism and Staff Burnout. We did not include burnout in the first model since it was assessed using a single item from the healthcare worker survey, whose validation with a more accepted Maslach Burnout Inventory has been conducted outside of resource-limited setting [23]. Similarly, we did not include absenteeism in the first model since it itself can be considered as an outcome of other staff related variables and its direct impact on health outcomes was, a priori, not clear. Since the results were not very different for the two model variants and because the coefficients of absenteeism and burnout were significant, we present the results of Model 2. All analyses performed using SAS/STAT software, Version 9.1 (Cary, NC, USA).
