Absenteeism & HIV/AIDS: A hospitality industry case study

March 2006

Neil M. Orr (MA Res.Psych, UCT)

David R. Patient (MHT)

1. Introduction

This report concerns the impact of HIV and AIDS on absenteeism in a hospitality industry company in Gauteng. Sick leave is probably the most direct measure of the impact of illness and health at the workplace. However, family responsibility leave also has an impact upon work, but reflects instead the impact of the health of employee’s families.  If there are changes in illness patterns over time – such as due to HIV and AIDS – it would be seen in both sets of data.

This article follows a specific line of questioning regarding the health (and illness) status of employees in the company: Have levels of illness changed over time? How much? What has caused such changes? What does this mean for the future?

2. Data Description

The sick leave (SL) data analysed excluded: Family responsibility leave, maternity and paternity leave; injuries at work, study leave, and any other form of leave. Similarly, family responsibility leave (FRL) refers only to this category of leave.

The core business of the company that commissioned this study is hospitality. The data analysed covered a 76 month period, from July 1998 to October 2004, at the company’s Gauteng unit. The data included 8,573 cases (20,425 days) of sick leave (SL) – averaging 2.38 days per case, and 1353 cases (3,105 days total) of family responsibility leave (FRL), averaging 2.29 days per case. The average headcount from 1998 to 2004 was 1,056.  The headcount varied little from year to year. There was a consistent ratio of 16 days of FRL per 100 days of SL.

The data was grouped according to whether the person was on medical aid (OMA) or not on medical aid (NMA), for costing purposes. The OMA to NMA ratio was 40:60. This ratio did not vary significantly from years to year.

The raw data was sorted according to whether each illness incident had a specific description (e.g., flu, injury, etc) indicated in the database (known), or not (unknown). Unfortunately, the recorded reasons provided for FRL were not specific enough to be used in a detailed analysis of causes.

Headcount and average monthly salary data – split into OMA and NMA – were entered into the analysis.  The average salary per month was obtained for each month of the period of analysis. The purpose of this data was to obtain a costing for SL and FRL over time. This article reports on percentage of payroll (e.g., sick leave as a % of payroll), not actual payroll numbers.

The average monthly salary was then divided by 21.67 working days to obtain the payroll value (cost to company) of one working day (OMA and NMA, separately). In order to present SL and FRL data in a meaningful manner, the number of days taken off from work was calculated as a percentage of the total possible working days (SL and FRL as %WD) for that same period of time. This allows direct comparisons over time, and calculation of costs-to-company for this leave taken.

3. Sick Leave & Family Responsibility Leave Trends 1998-2004

Figure 1 shows the trend in sick leave (SL) as a percentage of total working days from 1998 to 2004.

Figure 1 Sick Leave (SL) Days as % of Working Days (1998-2004)

From Figure 1 it is apparent that over a 6 year period commencing in 1998, sick leave (SL) - as a percentage of working days – has increased 528%, starting with 0.36% in 1998 and reaching 1.90% in 2004.  Personnel on medical aid are consistently above the annual average (2.05% of working days; increase of 512.5% from 1998), and those not on medical aid are slightly below the annual average (1.80% of working days; increase of 666.7% from 1998). In both groups, the increase has been parallel to the annual increase.

A similar trend exists for family responsibility leave (FRL), with a 517% increase in FRL from 1998 to 2004. On average, for every 100 days of sick leave, there were 16 days of FRL. When the two sets of data are combined, the overall increase is 527.2% from 1998 to 2004. Current levels (2004) of SL plus FRL were 2.21% of all working days for all personnel, while those on medical aid (OMA) were 2.37%, and NMA were 2.11%.

OMA personnel are typically management, and thus OMA leave days have a higher cost-to-company. One possible reason is that it is easier for someone on medical aid to seek medical attention, and thus they do so. The same was found for FRL, even though the average person on medical aid has a smaller family (average of 2.8 members in this population) compared to those not on medical aid (4.1 members).

It is clear that both sick Leave (SL) and family responsibility leave (FRL) has been increasing significantly and steadily over the 1998 to 2004 (6-year) period.

4. Causes for illness: 1998 to 2004

The number of SL cases that state the actual reason (i.e., illness) were too few for the 1998 to 2000 year to include in this analysis. Therefore, this analysis focuses exclusively on the 2001 to 2004 period.  The SL data was sorted according to the reasons stated on the sick notes submitted.  The results are summarized in Figure 2 below.

It is clear that almost 50% of all sick days are due to only three illness categories: RTI, GI, and ENT.

In order to explain the annual increases found in sick leave, we need to consider two obvious possibilities:

Hypothesis 1: There is no increase in real illness, only increases in sick leave days taken. I.e., abuse of sick leave, primarily 1-day sick leave. This possibility can be tested by comparing data that excludes versus includes all 1-day leave. In order to make such comparisons over time possible, the data was transformed into a case-per-thousand (per month) basis.

Hypothesis 2: There is an actual increase in specific illnesses types, possibly related to HIV.  The majority of personnel were between the ages of 20 and 35 years of age. Almost one third are between 26 and 30 years of age. This age group constitutes the highest HIV prevalence of all age groups (Dept, Health, antenatal figures for Gauteng, for age group 25-29, 1998 to 2003). If HIV is the primary cause for SL increases from 1998 to 2004, then common HIV-related pre-AIDS conditions would be escalating faster than non-HIV-related conditions.

An index for such common illnesses that increase in frequency in HIV-infected people was created, entitled P-Pos (for possible HIV-positive connection). The P-Pos index is not an indicator of who does and does not have HIV. Anyone can get a cold, flu or upset stomach. Illnesses with the least likely connection to HIV infection would be illnesses such as injuries (broken bones and motor vehicle accidents), backache, dental problems, and sinusitis. An index of such illnesses was created, entitled P-Neg (unlikely connection to being HIV-positive).

It is unlikely that the population would display marked presence of AIDS-specific illnesses. When a person develops AIDS illnesses, they are typically too ill to work, and placed on disability. ART had only been recently been introduced to those on medical aid, and uptake was quite low. Personnel not on medical aid did not have access to ART at the time of the study. 

The results are illustrated in Figure 3 below.

There are two things that are clear from Figure 3:

(a) Excluding 1-day sick leave data made no difference to the trends for both P-Pos and P-Neg groups of illnesses. I.e., whatever is driving the increases in sick leave levels has little to do with the abuse of 1-day sick leave.  

(b) P-Pos illnesses are in fact the main driving force behind annual increases in sick leave. P-Neg illnesses hardly change in frequency over time. In fact, it is clear from Figure 3 that the P-Pos category of illnesses – mainly RTI and GE – have doubled in incidence from 2001 to 2004 (100.2% with 1-day SL included; 89.9% with 1-day SL excluded). In contrast, the P-Neg category of illnesses has increased comparatively much less (34.6% with 1-day SL included; 30.5% without 1-day SL). In fact, the incidence in 2002 (21.7 cases/1000 per month) is almost identical to the incidence in 2004 (21.8). Refer P-Neg lines.

A closer examination of the specific illnesses in the P-Pos category (GE and RTI) found that both types of illnesses have increase in frequency: GE (especially gastro-enteritis) has increased 77.7% in incidence. In RTI (flu, bronchitis, pneumonia), the incidence increased by 112.8% from 2001 to 2004.

4.4  Statistical evidence of SL/HIV correlation

Although there is some evidence that the increase in sick leave is related to HIV, it is necessary to establish what these levels are, and what this means in terms of sick leave levels. This will allow us to establish future trends. The first factor is the clinical nature of HIV infection:

Figure 4

Generalized time course of HIV infection and disease

(Intermediate / Average Progressor)

Modified from: Centers for Disease Control and Prevention. Report of the NIH Panel to Define Principles of Therapy of HIV Infection and Guidelines for the Use of Antiretroviral Agents in HIV-Infected Adults and Adolescents. MMWR 1998;47(No. RR-5):Figure 1, page 34. Epidemiology of Disease Progression in HIV / HIV InSite Knowledge Base Chapter / Published May 1998. Dennis H. Osmond, PhD, University of California San Francisco; http://hivinsite.ucsf.edu/InSite?page=kb-03-01-04

Figure 4 represents the average clinical progression from HIV infection to AIDS. In reality, individuals vary in how fast or slow this progression occurs. The variance in such progression forms a Bell curve, with the profile in Figure 4 as the median.

Based upon the clinical profile described in Figure 4, a typical HIV-infected staff member will experience illness at different times during his or her infection:

(a) Flu-like symptoms during seroconversion (i.e., immediately after infection). These symptoms are mild, such as seroconversion rash, or mild flu-like symptoms. Many people do not experience or notice symptoms at this stage.

(b) Relative Health: CD4 count above 350. Seroconversion is typically followed by several years of relative health, even although CD4 (Helper T-cell) counts slowly decline, and HIV viral load climbs steadily. The T-cell (CD4) count is adequate (above 350 to 400) to ward off most illnesses. 

(c) Vulnerable: CD4 count 350 to 200. Once the T-cell (CD4) count reaches about 350 – typically after 6 years from infection - the immune system –although strong enough to ward off serious life-threatening illnesses, is weak. Therefore, common (non-life-threatening) illnesses become more frequent. Examples of such illnesses are recurrent upset stomachs (gastro-intestinal), and recurrent infections, such as herpes, thrush (candidiasis), shingles and respiratory tract infections. The frequency of these illnesses increases until year 8, namely the onset of AIDS illnesses.

(d) AIDS: CD4 count 200 or below. Once the T-cell (CD4) count drops below 200, AIDS symptoms (life-threatening) appear, and the staff member should probably be placed on antiretroviral (ART) treatment.  In the absence of ART, disability ensues. Life expectancy – without treatment drops significantly from year 8 to 10, with most (untreated) people dying by year 11.

The previous paragraphs describe the average progression of HIV to AIDS. There are indeed anomalies and variations, such as development of AIDS-defining illnesses prior to CD4 levels reaching 200, and also the absence of illness in those with CD4 levels below 200.

Therefore, when looking at the sick leave cases for a specific year, these cases should statistically reflect the following components:

1. Illnesses, accidents and injuries experienced by HIV-negative staff. Although there may be an increase in stress-related illnesses in this sector, these increases should remain relatively stable year-on-year. According to the 6-year infection-to-350-CD4 clinical progression model, 1998 sick leave would reflect HIV infections from 1992 and earlier. In 1992, this figure was 2.2% nationally, and 2.5% in Gauteng.  Therefore, if there is a hypothetical ‘in the absence of HIV’ sick leave level, this would be lower than the 1998 SL level of 0.36%.

2. There should be a significant increase in the incidence of colds, flu, and gastro-intestinal incidents after 6.years of infection. I.e., from year 7 onward.

3. There should be highly elevated levels of illness in those infected after year 8, (i.e., from year 9) followed by a drop-off as people develop AIDS and exit the work-force (unless treated with ART).

In order to determine whether sick leave is indeed driven by this specific clinical pattern, a Pearson’s correlation coefficient (measure of linear correlation between independent and dependant variables) was obtained between sick leave levels (1998-2004) as the dependant variable, and the number of people infected (in each year) for 1 to 11 years. To do this, the total prevalence rates from each year were deconstructed into new infections, deaths, and the percentages of infected people infected for 1 to 11 years.

Year 1 consists of new infections from the same year as the SL data. Year 2 is sick leave correlated with new infections from the previous year. I.e., Year 2 reflects new infections occurring 12 to 24 months previously. For 1998 and 1999, the correlation was performed for those infected up to 8 and 9 years, respectively. The results are described in Table 1.

Table 1

Sick Leave (SL) correlations with Years Infected

The correlations clearly show that sick leave is indeed linked to HIV infection, as it follows the expected associations with clinical stages of infection.  I.e., the more people who have been infected longer than 6 years, the higher the illnesses experienced in sick leave incidents, per year.

There is an apparent anomalous correlation for year 2 (infected between 12 and 24 months). This reflects the period after seroprevalence when the viral load is at its’ lowest level during the entire infection period.  A similar phenomenon sometimes occurs when the viral load is initially driven down by antiretroviral therapy (ART), called IRIS (Immune Reconstitution Inflammatory Syndrome). We cannot determine the reason for this correlation.

What is important to note from the model, is that current infection rates have their greatest impact 7 to 11 years later. I.e., current (2004) increases in sick leave levels are the result of infections occurring in 1993 to 1997. The implications are this: Even if new HIV infections were brought to a total and immediate halt, there will be an increase in sick leave for at least a further 11 years. This is due to the fact that those infected 7 to 11 years earlier still need to go through the clinical cycle.

It is worth pointing out that until the 6^th year of infection, there is hardlyany impact at all on sick leave levels. Efforts to extend this period of healthy time – e.g., wellness programs - would be well worth the effort. 

It should also be noted that there is a one-year difference between the clinical profile of the unit’s personnel and the average expected clinical progression profile. This may be due to the fact that the personnel in this company are slightly healthier than the normative group used to develop the average clinical profile (refer Figure 4), possibly due to factors such as quality nutrition in staff canteens, and access to good medical treatment for some personnel. Previous surveys indicated that at least one-third of personnel’s nutritional consumption occurred in the staff canteen. Efforts to further improve nutrition and medical care were underway in 2005.