Table 2. Details of parameter estimation, level of stratification, and data sources.

Model Input or Assumption	Level of Stratification	Source of Data (2003–2010)	Method of Estimation
			2003–2010 1	2011–2022
1. Changes in workforce
Workforce needed at end of year	Job grade, year	Company data	Data taken as is to calculate number of recruits or retrenchments required	Assumed to remain same as in 2010
Number of recruits	Job grade, year		Set to produce workforce needed at end of year	Same as for 2003–2010
Prevalence of recruits/retrenchees 2	Job grade, gender, year (for retrenchees, also by age)	Company data	N: all new employees with a positive first HIV result in the year of recruitment; D: all new employees with a positive or negative first HIV result in the year of recruitment	Assumed to remain same as in 2010
Distribution of recruits	Age group, gender, year	Company data (distribution set to be same as workforce distribution in database in 2003–2010)	N: number of employees in database by year, job grade, gender, and age group; D: total number of employees across all job grades, age groups, and genders by year	Assumed same as average 2003–2010
Annual rate of promotion	Job grade, year	Company data for 2005/2006	Assumed to remain same as in 2005/2006	Assumed to remain same as in 2005/2006
2. Start population and coverage
Distribution of start population (all employees)	Age group, gender, job grade	Company data	Number of employees in database by 31 Dec 2002 by job grade, gender, and age group	N/A (start year only)
HIV status of start population (all employees)	HIV status of those employees with an HIV test	Company data	Number of HIV-positive employees tested before 31 Dec 2002 and assumptions regarding untested employees’ HIV status	N/A (start year only)
Distribution of start population into CD4 cell count categories (HIV-positive employees)	CD4 cell count category	No data	Same proportion assumed in each CD4 cell count strata	N/A (start year only)
Baseline HCT coverage 3	Age group, gender, job grade	Company data	Number of employees tested before 31 Dec 2002 by job grade, gender, and age group	N/A (start year only)
3. Costs
Average basic salary	Job grade	Company data (payroll)	Salaries in cost year (2006)	Real cost assumed constant over time
Incremental replacement cost for HIV-positive employees	Job grade	Interviews with company human resources department	Average cost per new employee by job grade in cost year (2006)	Real cost assumed constant over time
Number of years that benefits get paid	None	Company benefit policy	Company policy	Real cost assumed constant over time
Incremental inpatient/outpatient cost for HIV-positive employees in cost year (2006)	Type of care (ART/no ART), CD4 cell count category	Bottom-up cost analysis of company health services	Average cost per employee in cost year (2006); includes non-ARV drugs, non-ARV-specific laboratory tests, patient contact time, other medical supplies, site programme cost, but no central management cost	Real cost assumed constant over time
Annual per employee cost of ART in cost year (2006)	CD4 cell count category	Bottom-up cost analysis of company health services	Average cost per employee in cost year (2006); includes central management cost for ART programme, ARV drug cost, ART-specific laboratory tests (CD4, VL)	Real cost assumed constant over time
Incremental absenteeism cost for HIV-positive employees	Type of care (ART/no ART only), CD4 cell count category, job grade	Payroll data on sick leave days	Absent days/shifts lost to sickness (sick leave) by health state in cost year (2006) multiplied by job-grade-specific salary per day/shift	Real cost assumed constant over time
4. Transitions between CD4 cell count categories
Transition probabilities	Type of care, CD4 cell count category	No care: public sector data based on [52]; all else: company data	N: all employees with a CD4 cell count in one stratum in time period t who have a CD4 cell count in a different stratum in time period t + 1; D: all employees with a CD4 cell count in one stratum in time period t that also had a CD4 cell count in time period t + 1	Assumed constant over time
Transition probabilities	Type of care, CD4 cell count category	No care: public sector data based on [52]; all else: company data	N: all employees with a CD4 cell count in one stratum in time period t who have a CD4 cell count in a different stratum in time period t + 1; D: all employees with a CD4 cell count in one stratum in time period t that also had a CD4 cell count in time period t + 1	Assumed constant over time
5. HIV incidence; coverage with testing, care, and ART; and treatment failure and retention
Incidence	Job grade, CD4 cell count category 4 , year	Change in HIV incidence over time fitted to company data on HIV incidence [53]; job grade weights: company data; CD4 cell count category weights: assumed	HIV seroconversion was assumed to occur at the midpoint between the first positive and the last previous negative HIV test; N: all employees with a calculated seroconversion date in one year; D: all employees with a negative HIV result and no seroconversion date in the previous year. This analysis excludes employees whose HIV test result was given as “unknown”	Assumed same as average of 2008–2010
Coverage with HIV testing, wellness care, and ART	Type of care, year, and, for ART, also CD4 cell count category	Company data	Model fitted to reported proportions of HIV-positive employees in each type of care	Assumed same as average of 2008–2010, except transition to first-line ART from wellness care, which is used to achieve ~92% ART coverage of eligible population
Rate of treatment failure	Year (same for first- and second-line ART)	Company data	N: employees with a failure start date during time period t; D: all employees on ART at the beginning of time period t	Assumed same as average of 2008–2010
Loss-to-follow-up rate	Type of care, year	Company data	N: all employees with a care stop date (wellness care and ART only) during time period t; D: all employees in wellness care and ART, respectively, at the beginning of time period t	Assumed same as average of 2008–2010
6. Separation rates
HIV-related	Type of separation, CD4 cell count category	Company data	Ill-health, death, and other non-transfer separations were allocated to a CD4 cell count category using the last available CD4 cell count before exit from the workforce from the database; N: all HIV-positive employees with an employment stop date by separation category and CD4 cell count category; D: all employee-years in the same CD4 cell count category	Assumed constant over time
HIV-unrelated	Type of separation, job grade	Company data	N: all HIV-negative employees with an employment stop date by separation category and job grade; D: all employee-years in the same job grade	Assumed constant over time

“Company data” refers to the mine company’s employee database of 9,211 employees and a separate database documenting the 1,149 employees who tested HIV positive and were enrolled in the company’s HIV care programme. The databases cover the period January 2003 to December 2010.

¹Details of analysis are given if a parameter was analysed from the company’s employee database. D, denominator; N, numerator.

²If the workforce is set to be reduced during one year, the resulting number of recruits will be negative, signifying the number of people who will be retrenched, rather than recruited, during that year.

³Coverage with all other care is set to zero at baseline.

⁴Incidence is stratified by CD4 cell count category to allow the distribution of newly incident members of the infected population into CD4 cell count categories. The values of the weights are 0.1, 0.2, 0.3, 0.5, and 1 for the categories >350, 200–350, 100–199, 50–99, and <50 cells/mm³, respectively.

ARV, antiretroviral; N/A, not applicable.