Cost-effectiveness of Sick Leave Policies for Health Care Workers with Influenza-like Illness, Brazil, 2009

TOC Summary: Seven-day leave was more costly and no more effective than 2 days plus reevaluation.

We describe the effect of infl uenza-like illness (ILI) during the outbreak of pandemic (H1N1) 2009 on health care worker (HCW) absenteeism and compare the effectiveness and cost of 2 sick leave policies for HCWs with suspected infl uenza. We assessed initial 2-day sick leaves plus reassessment until the HCW was asymptomatic (2-day + reassessment policy), and initial 7-day sick leaves (7-day policy). Sick leaves peaked in August 2009: 3% of the workforce received leave for ILI. Costs during May-October reached R$798,051.87 (≈US $443,362). The 7-day policy led to a higher monthly rate of sick leave days per 100 HCWs than did the 2-day + reassessment policy (8.72 vs. 3.47 days/100 HCWs; p<0.0001) and resulted in higher costs (US $609 vs. US $1,128 per HCW on leave). ILI affected HCW absenteeism. The 7-day policy was more costly and not more effective in preventing transmission to patients than the 2-day + reassessment policy.  (3). In São Paulo, the most populated state in the country (≈40 million inhabitants), the reported incidence was higher than anywhere else in the country (15.17 cases/100,000 inhabitants); 479 persons died (4).
Hospital das Clínicas (HC), the largest hospital in Brazil, was assigned by the State Health Department as 1 of the reference hospitals for persons with severe pandemic (H1N1) 2009 in the city of São Paulo. During the pandemic, specifi c sick leave policies were instituted at HC for health care workers (HCWs) who had infl uenza.
Considerable concern exists among HCWs about the risks of working during an infl uenza epidemic. Although they feel responsible to care for patients, they also are concerned about their own and their families' health (5,6). Transmission of infl uenza from HCWs to patients under their care is also a concern (7). Isolation precautions needed to prevent transmission of pandemic (H1N1) 2009 virus were heavily debated, and recommendations from the US Centers for Disease Control and Prevention (Atlanta, GA, USA) and WHO confl icted (8). Guidance on appropriate sick leave policies to avoid transmission from HCWs varies and is not well established (9).
Interim guidelines for protecting HCWs from pandemic (H1N1) 2009 (9) suggest that HCWs in whom fever and respiratory symptoms develop should be excluded from work for at least 24 hours after defervescence. HCWs caring for severely immunocompromised patients should be reassigned or excluded from work for 7 days after symptom onset or until resolution of symptoms, whichever is longer.

Cost-effectiveness of Sick Leave Policies for Health Care Workers
with Infl uenza-like Illness, Brazil, 2009 Workforces at large tertiary care hospitals functioning as reference hospitals for persons with infl uenza may be substantially affected during pandemics, particularly in regard to absenteeism and associated costs. Effectiveness and costs of sick leave policies should be evaluated to guide hospital managers and public health offi cials, particularly during epidemics and pandemics.
The objectives of this study were to describe the effects of ILI during the pandemic (H1N1) 2009 outbreak on HCW absenteeism and the associated costs. Furthermore, we aimed to compare effectiveness and cost of 2 policies for HCW sick leave during the fi rst wave of pandemic (H1N1) 2009 in a large urban tertiary care hospital.

Location and Setting
The study was conducted at HC, in the city of São Paulo, São Paulo state, Brazil. São Paulo is among the largest cities in Latin America, with 11 million inhabitants. HC, a tertiary care teaching hospital complex affi liated with the University of São Paulo, is a government hospital predominantly for patients covered by the publicly funded Brazilian National Health Service. During the outbreak of pandemic (H1N1) 2009 in Brazil, HC was assigned by the State of São Paulo Health Department as a state reference hospital for persons with severe pandemic (H1N1) 2009. HC has ≈2,000 beds distributed in 7 institutes. The Central Institute (main building) has 894 beds, including 100 intensive care unit (ICU) beds. The other centers are the Heart Institute (444 beds); Orthopedics Institute (152 beds); Children's Institute (174 beds); Psychiatry and Neurosurgery Institute (90 beds); Cancer Institute (213 beds); and Radiology Institute, without inpatients, which serves all other institutes. Although also part of the HC complex, the administration building, 2 long-term care facilities (234 beds total), and 1 outpatient rehabilitation center were not included in this study.

Infl uenza Triage, Diagnosis, and Treatment
As recommended by international guidelines, triage for suspected infl uenza cases was put in place at all of HC´s institutes. Triage occurred in the existing emergency departments in the Central, Heart, and Children's Institutes. Persons with confi rmed infl uenza were then referred to designated units within HC for hospitalization. Included in the Central Institute were an ICU (7 beds), an infl uenzaspecifi c ward (11 beds for semi-intensive care), and a specifi c ward for pregnant infl uenza patients (15 beds). In the Children's Institute, an ICU (2 beds) , a ward (14 beds), and 2 rooms at the emergency department were assigned. Later into the infl uenza season (August 2009), 2 additional infl uenza-specifi c wards were designated in the Heart Institute, with 16 rooms and 32 beds. Although persons with infl uenza should have been hospitalized preferentially in the above-designated units, a few patients in whom infl uenza was only suspected after admission and who could not be transferred were hospitalized in nondesignated units.
Laboratory testing for confi rmatory diagnosis was conducted for all hospitalized persons suspected to have infl uenza. Rapid diagnostic tests for infl uenza were not performed during the pandemic. Before July 2, 2009, samples were collected and sent to the Adolfo Lutz Reference Laboratory, where real-time PCR (rt-PCR) testing was conducted. By July 2, 2009, HC Central Hospital Laboratory started the infl uenza A (H1N1) rt-PCR. Although available for patients admitted to the hospital, laboratory confi rmation of pandemic (H1N1) 2009 infection was not available for HCWs. Thus, suspected cases among HCWs were triaged according to the attending physician's evaluation. The rt-PCR for suspected infl uenza in HCWs was made available only on August 24, 2009. The PCR protocol in place was developed by the US Centers for Disease Control and Prevention (10).
During the entire pandemic period, active surveillance for nosocomial pandemic (H1N1) 2009 was performed by the infection control teams. A suspected infl uenza case was defi ned, according to national guidelines (11), as fever and respiratory symptoms such as cough or sore throat in the absence of other diagnoses, and confi rmed cases were defi ned as those with positive test results for pandemic (H1N1) 2009.

Infection Control Policies
Following global and national guidelines (8,12), infection control policies for prevention and control of nosocomial transmission of infl uenza were in place in the hospital. These included contact and droplet precautions taken by HCWs during the care of patients with suspected and confi rmed infl uenza and use of N95 masks only during aerosol-generating procedures.
HC has a centralized occupational health service for workers of all institutes. We assessed ILI in and ILIassociated sick leave for HCWs who are overseen by the hospital's occupational health service, i.e., physicians; nurses and nurse assistants; and pharmacy, nutrition, laboratory, and administration workers. Cleaning and security services are furnished by third-party service providers; therefore, their staff are not overseen by HC's occupational health service.
Starting on July 16, 2009, in HC institutes except the Heart Institute, an HCW with suspected infl uenza received an initial 2-day leave and was reassessed every 2 days until asymptomatic, when he or she returned to work (2day leave + reassessment policy). In the Heart Institute, a different sick leave policy was adopted in which an HCW with suspected pandemic (H1N1) 2009 infection received a 7-day leave (7-day leave policy). During the prepandemic period, sick leave for respiratory infection was determined on a case-by-case basis after evaluation by a physician (Figure).

Study Design
We retrospectively evaluated the effect of ILI on HCW absenteeism during the pandemic (H1N1) 2009 outbreak. The main outcome of interest was HCW absenteeism.
We estimated the economic effects of HCW-associated illness from the hospital´s perspective. Costs were estimated retrospectively. Among the various techniques for measuring costs, 1 of the most commonly used is the accounting approach (the conventional costing method). This approach can be divided into 2 categories: the fi rst uses detailed, bottom-up, step-down analyses of accounting to distribute shared costs across the activities considered (also called ingredients approach). The second uses a top-down approach, which makes less detailed estimates of highlevel average costs on the basis of aggregate expenditure records. In this study, we used the bottom-up approach as we considered individual ingredients to estimate costs. We compared the effectiveness and cost of the 2 sick leave policies: initial 2-day + reassessment policy and 7-day.

Study Period
The study was conducted during the fi rst wave of pandemic (H1N1) 2009 in the Southern Hemisphere winter season (May-October 2009). The prepandemic period was defi ned as August 2008-April 2009. HCW was defi ned as a professional on the hospital's payroll, even if the worker did not work directly with patients. This defi nition excluded students and residents/fellows, as well as workers of thirdparty service providers (cleaning and security).

Data Analysis
We calculated the monthly number, average duration, and rate of infl uenza sick leaves issued in HC by institute and HCW profession. We determined the proportion of sick leaves taken for ILI in relation to the total number of sick leaves during the pandemic (H1N1) 2009 outbreak (May-October). Rate of infl uenza-associated sick leaves issued during the pandemic (H1N1) 2009 outbreak (August-October 2009) were compared with the prepandemic period (August-October 2008).
To assess the effectiveness of the 2 sick leave policies, we evaluated the number of nosocomial pandemic (H1N1) 2009 cases. We calculated the total number of sick leaves issued, their average duration, and total days of ILIassociated sick leave for HCWs working under the 7-day and 2-day + reassessment policies during the pandemic period. In addition, monthly rates of ILI-associated sick leave days per 100 HCWs were compared for the 2 sick leave policies by using χ 2 .

Cost Analysis
The cost analysis considered only ILI and absenteeism in physicians, nurses, and nurse assistants because of the availability of information. Direct medical costs of diagnosis, treatment, and follow-up for HCWs with suspected and confi rmed pandemic (H1N1) 2009 were estimated. Direct medical costs also included the cost of replacement HCWs during sick leave. Direct nonmedical costs assessed were transportation of symptomatic HCWs to and from HC for assessment. Indirect costs comprised productivity losses of HCWs during sick leave. Costs for outpatient treatment of HCWs suspected to have infl uenza comprised medical assessment by the occupational health clinic, specimen collection and rt-PCR (only after August 24, 2009) for laboratory diagnosis, and treatment. We assumed that 100% of HCWs with suspected infl uenza received antiviral medication (oseltamivir) according to the offi cial hospital guideline. Twenty percent received antimicrobial drugs (amoxicillin) for secondary bacterial infection, and 70% received symptomatic medication.
HCW hospitalization costs associated with pandemic (H1N1) 2009 were estimated by considering the average When evaluating cost of staff replacement, we assumed that 70% of nurses and nurse assistants and 50% of physicians on sick leave were replaced. We also assumed that HCWs under the 2-day + reassessment policy would return on average 1.5 times for reassessment, thus incurring medical consultations and transportation costs for 1.5 return visits to HC. Under the 7-day policy, we assumed that 10% of HCWs would return for medical consultations during leave; thus, only 10% of HCWs required transportation for reassessment.
We estimated productivity losses by considering the daily 12-hour average salaries for each HCW category based on an average of 5 years of work in HC. Wages and mandatory Social Security and health care contributions were considered. Monthly salary was Brazilian real (R$) 3,071.54 for nurses working 30 h/week; R$2,541.55 for physicians working 20 h/week; and R$1,571.80 for nurse assistants working 30 h/week. One US dollar was equivalent to ≈R$1.80.
We also recalculated the costs for each sick leave policy on the basis of a lowest cost scenario and a highest cost scenario. For the lowest cost scenario, we assumed that HCWs on leave would have returned on average 0.5 times for reevaluation if under the 2-day + reassessment policy and would not have returned for the 7-day policy. We also assumed that 5% of HCWs received antimicrobial drugs and 20% received treatment for symptoms. Staff replacement was assumed to have been 30% for nurses and nurse assistants and 10% for physicians. For the highest cost scenario, we assumed that each HCW had 3 return consultations for the 2-day + reassessment policy and 1.5 return visits for the 7-day policy, that 60% of HCWs received antimicrobial drugs, and that 100% received treatment for symptoms. Staff replacement in the highest cost scenario was assumed to be 100% for nurses and nurse assistants and 80% for physicians.

Data Sources
The inputs, values, and data sources are presented in Table 1. HC maintains a centralized occupational health database in which sick leave of HCWs since August 2008 is registered by International Statistical Classifi cation of Diseases and Related Health Problems (13). The number of HCWs placed on sick leave during August 2008-October 2009 because of diseases of the respiratory system reported as codes J10 (Infl uenza due to other identifi ed infl uenza virus) and J11 (Infl uenza, virus not identifi ed), by using the International Statistical Classifi cation of Diseases and Related Health Problems, 10th Revision, was evaluated (13). Cost data were obtained from the hospital administration.

Results
During June-September 2009, a total of 796 persons with suspected infl uenza were hospitalized at HC, for which 214 infections were laboratory confi rmed as pandemic (H1N1) 2009 (2). In July 2009, HCW sick leaves began to increase and peaked in August, when 3% of the workforce received leave for ILI (Table 2). HCWs received 884 ILI-associated sick leaves during August-October 2009, compared with 96 during the same period in 2008 (p<0.00001). Of 244 HCWs tested for pandemic (H1N1) 2009, a total of 52 (21%) received positive results. The mean monthly rate of infl uenza sick leaves per 100 HCWs was signifi cantly higher in the Heart Institute, which had a different sick-leave policy from the other institutes (p<0.0001) ( Table 3). The distribution of ILI-associated sick leave of HCWs varied by professional category (Table 4).
Three HCWs were hospitalized in HC because of pandemic (H1N1) 2009, resulting in a total of 34 days of hospitalization. One HCW was hospitalized in a private hospital; related costs were covered by private medical insurance and thus not included in our analysis. Total cost in all HC institutes was R$798,051.87 (≈US $443,362). At the Heart Institute (7-day sick leave policy), 169 staff (physicians, nurses, and nurse assistants) received sick leave because of ILI, resulting in a total cost of R$343,082.94 (≈US $190,602). At the remaining 6 institutes (2-day + reassessment policy), a total of 415 staff (physicians, nurses and nurse assistants) received ILI-associated sick leaves, resulting in a total cost of R$454,968.92 (≈US $252,761) ( Table 5). Thus, for each HCW on leave, cost was R$1,096.31 (≈US $609.06) for the 2-day + reassessment policy and R$2,030.08 (≈US $1,127.82) for the 7-day policy ( Table 6).
In the lowest cost scenario, total cost was R$617,135.45 (≈US $342,853.02). In the highest cost scenario, total cost was R$942,588.49 (≈US $523,660.27).
During the study period, active surveillance was conducted for hospital-acquired pandemic (H1N1) 2009 infections. No cases of infl uenza were documented or suspected in patients.

Discussion
Pandemic (H1N1) 2009 substantially affected HCW absenteeism with signifi cantly higher ILI-associated sick leave during August-October 2009 than during the same period in 2008. The 2-day + reassessment policy was less costly (approximately half the cost per HCW) and as effective for preventing transmission of pandemic (H1N1) 1426 Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 17, No. 8, August 2011  2009 as the 7-day policy because no cases of nosocomial acquisition occurred among patients. The effect on society of infl uenza leading to absenteeism is well documented. A systematic review of studies in this fi eld showed a loss of 1.5-4.9 workdays per episode of laboratory-confi rmed infl uenza (14), but this review did not focus on HCWs. The impact of infl uenza epidemics on sickness-associated absence in hospitals is diffi cult to defi ne. One study involving a hospital in the United Kingdom during 2 infl uenza seasons (1993-94 and 1996-97) showed that although the number of ILIassociated medical consultations in the population increased markedly during the outbreaks, they did not affect absences from work of hospital staff (15). Contrary to that study, our study demonstrated an effect of the pandemic: the number of HCW sick leaves increased greatly during the pandemic (H1N1) outbreak.
The lack of evidence-based policies for HCW sick leaves during infl uenza pandemics led us initially to consider that recommendations for patients (9) applied to HCWs, based on potential transmission for 7 days or even longer if symptoms persist. If we had followed this policy, our workforce would have been substantially reduced; thus our capacity to respond adequately to the pandemic would have been diminished. Because of that, we decided to apply the 2-day + reassessment policy. One of the HC Institutes did not comply with the offi cial policy, which allowed us to compare the policies. Active surveillance for hospital-acquired infl uenza among patients was essential to validate and allow continued reassessment of the policy. Although costs and cost savings may seem moderate by some standards, it is necessary to remember that the minimum monthly salary in Brazil is R$510, equivalent to ≈US $283.  Because diagnostic testing was not available for HCWs during most of the season, most of the decisions regarding sick leave were based on clinical evaluation, and we could not evaluate whether absenteeism resulted from infl uenza infections themselves or from other factors, such as the psychological impact of the pandemic and concerns about self and family health related to occupational exposure at work. A few studies have evaluated the effect a pandemic of respiratory disease might have on the attitudes of HCWs (16,17) and on the distribution of resources within hospitals (18). In Australia, only 50% of HCWs questioned said they would come to work during an avian infl uenza pandemic (16), and only 25% said they believed that their department was prepared to handle an infl uenza pandemic (17). In our hospital, diagnostic testing (rt-PCR) was available for HCWs at the end of the season, and only 21% of HCWs tested received positive results. Whether this percentage refl ects the previous period, when the test was not available, is not known.
These issues suggest that to mitigate concerns of HCWs, clear infection control strategies and hospital policies to protect HCWs and immediate testing of symptomatic HCWs and patients would reassure the workforce. Thus advanced planning and preparedness for implementation of such policies during epidemics and pandemics is needed.
Our study had several limitations. Part of the costs had to be estimated because of lack of data. We tried to counteract this limitation by estimating highest and lowest cost scenarios. In addition, diagnostic testing was available for HCWs only at the end of the season.
Some positive aspects were observed during the pandemic (H1N1) 2009 outbreak. Awareness increased that HCWs should not work when sick with respiratory infections. Usually HCWs tend to underestimate their health problems and work when ill, placing patients at risk (7). Another positive effect was the widespread use of alcohol-gel solutions within and outside the health care environment to halt transmission of infection. The pandemic presented an unprecedented opportunity for education of HCWs, children, and the general public about hand hygiene.
In conclusion, our retrospective study evaluated the effect on HCWs and associated costs of different sick leave policies implemented during the pandemic (H1N1) 2009 outbreak. The 7-day policy was more costly but not any more effective than the 2-day + reassessment policy in preventing transmission to patients. Decisions about HCW sick leave policies during pandemics should account for multiple factors, including effectiveness, cost, and feasibility of implementation during emergency conditions. Dr Mota is a medical administrator and coordinator of occupational health services in HC, University of São Paulo, Brazil. Her main research interest is quality assurance, especially as it relates to the health of HCWs.