We produced a demand-driven staffing model using simple spreadsheet technology, based on treatment protocols for HIV-positive patients that adhere to Mozambican guidelines. As such, it represents the minimum requirements to successfully complete protocols for HIV treatment and accounts for some, but not all, the extra encounters that could be generated by complications such as opportunistic infections. The user can easily adjust for the volumes of patients at differing stages of their disease, varying provider productivity, proportion who are pregnant, attrition rates, and other variables.
We relied on the document, "Human Capacity Development Assessment and Strategy Development for the Health Sector in Mozambique," previously referenced, and prepared by the Africa Bureau of USAID for some of our assumptions. This document will be referred to as the 'USAID document.' We also relied on the personal knowledge of four authors: Mark Micek, a physician with Health Alliance International (HAI), who was involved in the implementation of public-sector HIV treatment clinics in Beira and Chimoio, Mozambique; Kenneth Gimbel-Sherr, who is HAI's Mozambique country director and was involved in developing the original national plan with the Ministry of Health for providing HIV treatment nationwide; Ferruccio Vio, who works as Maputo Technical Support Coordinator for HAI; and Pablo Montoya, Central Mozambique Field Director for HAI, where he supports provincial planning for the Ministry of Health.
We assumed patients present to the health care system following a referral from one of a number of HIV testing sites within a community, including VCT (voluntary counselling and testing) centres, PMTCT (prevention of maternal to child transmission) centres, and hospitals, so they are known to be HIV-positive upon arrival. Patient CD4 (Cluster of Differentiation 4) counts, however, are unknown at the time of presentation.
Our model allows the user to input the distribution of patients eligible for ART at their initial contact with the clinic. We assumed that 45% of the HIV-positive adult patients will need to be placed on anti-retrovirals upon presentation, and that another 5% of HIV-positive presenters will be pregnant women who would also benefit from immediate ART treatment. For those not initially eligible for ART, we estimated that an additional 20% would have initial CD4 counts from 200–349/mm3, 15% would have CD4 counts between 350–499/mm3, and 15% would have CD4 counts ≥500/mm3. These estimations are based on the experience at Beira and Chimoio, where it is notable that the proportion of people needing ART exceeds UNAIDS and strategic plan assumptions. Patients presenting for HIV care may not represent a cross-section of all HIV-positive patients, but rather those who sought testing and successfully presented for care at an HIV treatment facility referral. The population who seek testing may be in the more advanced stages of illness than those who postpone. It should be noted, however, that the model allows these assumptions to be changed depending on differing experiences at different sites. At this stage, we are not including children in the analysis.
Our demand-driven model also provides the user with an opportunity to input a variety of additional assumptions, including 1) the proportion of those eligible for ART who would start treatment (we assumed 70%), 2) the proportion who will leave the care system secondary to death or loss to follow-up (we assumed 10% per year for those on ART, 50% for those enrolled but not eligible for ART), 3) those with adverse drug reactions (we assumed 10%), and 4) those who experience a lack of clinical improvement and therefore may require more encounters (we assumed 10%).
We assumed patients start ART according to Mozambique Ministry of Health guidelines, which include all patients with CD4 levels under 200 cells/mm3 regardless of clinical stage, a CD4 level between 200 and 349 cells/mm3 if also in WHO stage 3 or pregnant, or WHO stage 4 regardless of CD4 count.
Schedule of encounters
Our approach was to identify several 'types' of patients, and to map out the appropriate schedule of encounters for each newly-presenting type of patient based on published Mozambican guidelines .
'Encounters' in our model are from the care provider's point of view. A single patient trip to the clinic could generate several encounters if the patient sees more than one provider type during that trip. We will distinguish, therefore, between trips and encounters.
The first two patient trips consist primarily of assessment and planning procedures (including obtaining CD4 counts), so these would be the same for everyone. Trip 1 generates one encounter with a nurse for a clinical evaluation, and a separate encounter with a nurse who does a blood draw. This sample is sent to the lab, which takes 7 days to process and receive results.
Trip 2 is a week later than the first, at which time there is a single encounter with a nurse who evaluates the result of the CD4 count and makes a staging decision about the progress of the disease. This places the patient in one of several categories based on initial CD4 counts and clinical staging and on estimations about the rates in which people may change clinical categories over the duration of the 3 years of the model.
The patients who would need ART immediately would have an accelerated schedule of encounters: trip 3 would be within a week of the second trip and would generate two encounters: one with a clinician authorized to prescribe ART (mid-level medical technicians have been authorized to prescribe ART since June of 2006), and an encounter with a social worker to review the care plan. For these patients, trip 4 is with a social worker, as Mozambique recommends three encounters with a counsellor before ART is initiated. At trip 5, the patient is started on ART. There are encounters with the social worker, pharmacist and a clinician to discuss how the drugs will be administered and how to take them. At trip 6, two weeks after starting ART, there are encounters with a phlebotomist for haemoglobin and liver tests, a clinician, a pharmacist, and a counsellor to assess the course of therapy and review blood work results. Subsequently, these patients have monthly encounters to a pharmacist, and will see a clinician and counsellor at months one, two, four, seven and ten after starting ART. For pregnant women, encounters with a phlebotomist (for haemoglobin) and a clinician are also required at six weeks to monitor the side effects of AZT. Routine CD4 counts at month four and every six months thereafter require a trip and a encounter for blood draws. For each cohort starting ART, we estimate that 10% will have significant reactions or illnesses during the initial two months of treatment that will require further clinical encounters. In addition, at each CD4 draw time, we estimate 10% of patients will be identified as potential treatment failures, and will require additional encounters that are included in our model. Again, these assumptions are modifiable depending on differing experiences encountered at different sites.
For those patients who are not yet eligible for ART, we scheduled nursing encounters to repeat CD4 testing at intervals specified by Mozambique recommendations. This includes encounters every three months for those whose CD4 counts are between 200 and 349; encounters every 6 months for those with CD4 counts between 350 and 499, and encounters every 12 months for those with CD counts at or above 500.
To estimate encounters in the second and third years of follow-up for patients not initially ART-eligible, we estimated the proportion of patients presenting in each of the of clinical stages, and the rates at which these people may change clinical categories over the duration of the 3 years of the model. The encounter schedule will change, therefore, based on the progressing clinical stage.
The USAID document (p. 21) states that physicians can be expected to work 1600 hours per year, or 200 working days at eight hours per day. This assumes about 40 weeks of work a year, or significantly below typical U.S. working expectation of 45 to 48 weeks. Furthermore, the document discounts those 1600 hours by an additional 20% (ostensibly, but not explicitly, for administrative time) to yield 1280 patient contact hours per year. With 1280 contact hours per year over 200 days, or 6.4 hours per day, we calculate an encounter takes 12 minutes and that a clinician can be assigned roughly 6000 encounters per year.
Our model assumes nurses can be assigned 6000 encounters per year. A social worker would be able to complete 3000 encounters, and a pharmacist could process 10 000. Peer counsellors, or activists, are projected to be able to follow up five missing patients per day, and need is projected by multiplying 15% times the number of pharmacy encounters. A phlebotomist would be able to draw blood on 25 patients per day, or 5000 patients per year. Changing our productivity assumptions would change our staffing requirements, of course.
The same document estimates there were 647 physicians in Mozambique at the end of 2003, about 40% of whom were specialists. A draft Human Resources Development Plan calls for that number to more than double by 2010. This plan further calls for an additional 1255 nurses beyond the 4025 estimated to be practicing in 2004 .
In addition to the small numbers of personnel, other supply problems named include weak human resource management, too few administrative managers, low motivation levels of health workers, high turnover or loss of health workers secondary to HIV-related or other serious illness, and a shortage of protective equipment and supplies. Our report does not address these issues.
There are nine categories of health worker in our model. These include:
Adult non-obstetrical (non-OB) clinicians (physicians and ARV-trained mid-level medical technicians), trained to make decisions about ART therapy;
Adult non-OB clinicians, who can manage ART therapy, but are not trained or required to make decisions regarding starting or changing ART regimens, and do not need to be a physician or ART-trained mid-level medical technician;
Obstetrical clinicians, trained to both start ARTs and manage them through the patient's pregnancy;
Obstetrical clinicians who can manage ART therapy but are not trained or required to make decisions regarding starting or changing ART regimens;
Clinical nurses who can evaluate CD4 counts and make referrals to clinicians for ART therapy;
Phlebotomists, who can draw blood samples for CD4 and other blood tests and send them to laboratories for processing;
Social workers, who engage in pre- and post-ART counselling;
Pharmacists, who dispense ART drugs; and
Lay peer-counsellors (Activistas), who are an important component of the health care team but whose roles are not well defined. These individuals are responsible for finding patients who seem to be lost to follow up, and can also support chart management and receptionists and perform other all-around tasks, such as adherence counselling, patient orientation, and HIV prevention counselling.
Our model consists of four interconnected spreadsheets that build on each other but are relatively simple to understand. See Additional File 1 for a live spreadsheet workbook.
The fundamental spreadsheet (Worksheet A) uses the individual patient as the unit of analysis. It has 36 columns – one for each month of 3 years. The rows are grouped into five categories of patient type based on their health status at the time of enrolment: 1) those who need ART now, 2) those with a low CD4 count (200–349) who will need ART 'later' (within 1 year after enrolment); 3) those with a low CD4 count (200–349) who will not need ART within 2 years, 4) those who have CD4 counts of between 350 to 499, and 5) those whose counts are at or above 500. These numbers are automatically generated based on the assumptions entered in the 'Assumptions' section of Worksheet D.
Worksheet A maps out the encounters described in schedule of encounters,' above, over a three-year period.
The next spreadsheet (Worksheet B) also has 36 columns (one for each month), but the rows consist of total patient encounter counts generated in the patient-level Worksheet A, depending on the number of patients entered into the system. These total counts are also grouped by category of patient. The patient attrition assumptions (entered by the user) are played out in this spreadsheet.
The third spreadsheet (Worksheet C) is a large one, and is from the care system's point of view. Monthly encounter counts generated by type of patient are totalled and scheduled over a three-year period.
The initial "input control" spreadsheet (on the same page as Worksheet D) allows users to enter assumptions about patient distribution characteristics as well as to select one of three patient volume scenarios. Method 1 allows the entry of a total number of people on ART in the system at three time points one, two and three years after 'time zero'. Method 2 allows the entry of a total number of people for care (whether or not they are on ART) in the system at the one, two and three year time points after 'time 0' (these can be at either the clinic or national level). Method 3 allows the entry of a number of patients enrolled in a care system per month (at either the clinic or national level), and is intended to represent the care system's flows during a steady state period. Only one method at a time may be used. At this time, the model looks forward for only a three-year period.
When patient volume inputs are entered, a 'Summary Table' on the worksheet calculates numbers of patients enrolled per month for each of the three years in the model.