Effectiveness a n d cost-benefit of an influenza vaccination program for health care workers

A

periode de m ala die evitee) mes ures par rappor t aux coCtls (m ateriel, te m ps des infirm ieres en s a nte a u trava il.temps d es employes du rant Ia vaccina tion el h eures p erdues en ra ison des effe ts inclesira bles) revelenl un b en efice n el de 39,23 $ pa r employe vaccin e .Les ana lyses de sen s ibilile ont monlre qu e l'efficacile du vaccin et l'a bsenleisme a tlribu e a Ia grippe e t aux effe ts indesira bles de Ia vacc ina tion con s titu a ient les facie urs les plus importan ls: avec un nombre de j ournees perc! u es pour effe ts incles irables 0 .01 3 journees pa r personne vaccinee et un e effica cile du vaccin de 70 % .on pe ut oblen ir des b enefices n e ts quand l'a b s enteisme dCt a Ia grippe es t egal ou superie ur a 0.5% des jou rnees p ayees durant Ia saison epiclem ique.D'apres les resulta ts .un progra mme de vaccina tion des tine au personnel hospitalier diminue a Ia fois Ia m orbiclile parmi le personne l e l les infections n osocomia les provoquees pa r le virus grippa l ch ez les pa tients: sa renta bilite n e Ue sera il m ajoree pa r une promotion active-s urtout a upres du personnel oeu vranl clans d es sec leurs a h a ul risqu e.
I NFL UENZA CONTINUES TO BE A MAJOR CAUSE OF hospita lization and morbidity a m on g high ris k persons ( 1 ,2).Influen za A virus infection recurs a lmos t a nnua lly.and nos ocomia l s pread of influ en za is well known to occur (3)(4)(5)(6)(7)(8)(9).Although n ever unequivocally d ocumented , th ese ou tbreaks a re thou ght to be perpetu a ted by s pread of the virus to s u sceptible hos pital sta ff.who tra n s mit the infection to pa tients (9).Lost produ ctivity is a major sequela of influen za in h eal thy a dults , with ra tes of a b senteeism a mong hospita l workers of up to 3 0% h a ving been noted (7) .
The effic acy of influen za vaccine is a pprox ima tely 70% in h ealthy individua ls when the vaccine strain m a tch es the epidemic s train (10) .It h as b een s u ggested tha t reduced pa tient morbidity a nd control of nosocomial outbreak s of influ en za could b e accomplished by immunization of hospita l personn el (9) .In order to protect pa tients who h a ve a high risk of s uffering s erious complications following influenza infection, vaccina tion of h ealth car e workers who h a ve extens ive contact with th es e pa tients is recommended (2) .However , d espite the ava ilability of efficaciou s vaccines kn own to be cost effective when a dminis te red to high risk pa ti e nts (11.1 2). it is widely accepted tha t the vaccine is underemployed .Immuniza tion of hospita l worke rs m ay be diffic ult a nd les s effective once a n outbreak h as b egun (9) ; this con cern highlights the n eed for inten s ified efforts prior to each influenza s eason.
The a uthors docume nted increased a b senteeis m a mong hospital staff at their in s titution during the influen za epidemic of 1980-81 , a nd recommended a n influe nza vaccina tion progra m for h ealth car e workers (13).Such a progra m was ins tituted in 1984, with th e influen za vaccine offer ed to employees a nnu a lly during the months of October a nd November by the ins titution's occupa tiona l h ealth d epa rtment.All h ealth care worke rs we re conside red to be a t increased risk for acquis ition of influen za a nd wer e thus offer ed th e vaccina tion at no cost.Du e to the la rge employee popula tion a t this teaching hospita l (approximately 5500) , only the 8 00 employees working in areas d esign a ted as high risk (by virtue of the patient popu la tion and the extent of exposure) (2) received fo cused leaching a nd vaccine promotion.These a reas (Ta ble 1) wer e vis ited twice by an occupationa l health nurse who posted a n information sheet d escribing the va ccine; h er availability to return to a n s wer qu estions a nd to a dmin ister the va ccine was also discussed with the health care workers pres ent.Individuals were instructed to m a ke their own appointments to receive the vaccine .
Th e purpose of the present s tudy was: to review acce ptance of the existing h ealth care worker vaccination progra m; to a scertain if excess absenteeis m during the influ enza s eason could a gain be documented; to a ssess the extent to which influ en za season a b senteeism was r educed a mong those vaccina t ed; a nd to estima te the cost-ben efit of this progra m , a nalyzing the fac tors that most influen ced the results .

SUBJECTS AND METHODS
Employee charts in the occupational h ealth d epa rtment were reviewed to a s certa in the ra te of vaccine accepta n ce for the influenza s easons of 1984-85 to 1988-89 a s a whole and among health car e workers d esign a ted at increas ed risk (by virtue of pa tient popula tion and extent of patient contact, according to the Centers for Disease Control [CDC] guidelines [2]).The extent of ongoing participa tion in the vaccination program among those vaccinated wa s a lso assessed.
The lime period of the influenza season wa s d efin ed by a review of the number and types of influen za strains is ola ted from specimens (mainly throa t s wa b s ) routinely s ent from throughout th e province to the centralized virus d e tection labora tory of Cadha m Provincial La bora tory 1986-89.
The is ola tes were identified in the usual manner (1 3) with reagents provided by the La boratory Centre for Disease Control (LCDC) in Ottawa: at least on e isola te wa s s ent annually for r efer en ce typing to th e LCDC a nd to th e CDC in Atlanta, Georgia .In 1986-87 th er e were only three isola tes of influen za A virus (influen za AITaiwa n/1186like virus [H1N1]) documen ted .In contr a s t, 12 isolates of influen za AISichuanl 2 18 7 -like virus (H3 N2) a nd 29 isolates of influen za B intermedia te between influenza BlAnn Arbor l l/86-like virus and BIVictoria l2187-like virus were documented between J a nuary 1 and March 2 4, 1988 (Figure 1).From within the authors' institution ther e was no influen za isolate in 1986-87, compa r ed to 14 is olated from pa rticipants and staff in the 1987-88 season (five influenza A a nd nine influen za B).
Th er e was, therefore , only a partial m a tch b etween the vaccine offer ed in the a u tumn of 198 7 (which con tain ed influenza AILeningra dl360186 a nd B I Ann Arbor I 1 I 86) and the stra ins of influen za virus isola ted from persons in the hos pita l and the community in the 1987-88 winter season.Records for 1986-88 wer e availa ble to a s certain a b sen teeism ra tes (hours lost du e to s icknes s per hours pa id).The 1987-88 year was chosen a s th e focus of study .The rate of hours lost per hours pa id in the 14 week period during th e influen za season was compared to the eight week period before th e outb reak, the eight weeks a fter th e outbreak , a nd the rest of the calenda r year.With few isola t es a nd no s ignificant morbidity in the community during the 1986-87 winte r s eason , this year was used as a 'compa ris on' year (Figure 1) .
Of th e 92 person s who r eceived the vaccine in the fall of 1987, 52 wer e employed in a reas cons idered to b e a t h igh risk for acquisition of influenza.Of th ese 52, pre-vaccina tion abse nteeism data were availa ble for all bu t two employees ; s imila rly, post vaccina tion data wer e available for all but two The cost-benefit analysis was lim ited to consideration of work time lost re lated to influenza and/or the prevention program in addition to the direct costs of the program .Positive a nd negative outcomes which did not affect time lost from work were not considered.Thus, the important reason for vaccinating health care workers .ie , the potential benefit to hospital patients through reduction of nosocomial infection and its complications.was not considered.nor was the cost associated with responding to influenza outbreaks.Because the vaccination program was a small •add on' to the activities of the occupational health department, overhead and capital costs were not considered.Discounting of costs and benefits was not necessary.as they were accrued within the year of vaccination.
Benefits of the influenza vaccination program were measured as the dollar value of sick time avoided by the prevention of influenza, estimated from the absenteeism data.The costs of the vaccination program included: cost of time lost by recipients due to vaccination, and material and labour costs associated with the promotion and administration of the vaccine.Time lost due to adverse reactions was estimated from information voluntarily reported to the occupational health clinic.and thus represented a m inimum estimate.The average value of labour time was derived from the salary scale of general duty nurses, who comprise the largest group of h igh risk health care workers .0.037 0.003 post 'flu ' : 0.036 0.001 all o ther: 0.037 0.002 received vaccine in all four years of the program.The maximum proportion of repeat vaccinees was 41 % , which occurred in 1988; only ll% of th e 112 employees vaccinated that year received vaccine for three or more consecutive years.With an annual staff turnover of less than 5% in this group of health care workers, the attrition rate from the vaccination program is quite high.Absent eeism during influenza epidemics: As shown in Table 2, there was a s ignificant difference between absenteeism rates during the 1987-88 influenza season and the three noninfluenza periods in that year, with an absenteeism rate approximately 35% higher than either the pre-influenza.post influenza or remaining periods ofthe year.During the corresponding calendar period of the previous year, in which no influenza epidemic occurred.no increase in absenteeism occurred.

Vaccine acceptance:
The increase in absenteeism for employees of other areas of the hospital was not as great as among the 800 health care workers in h igh risk patient care areas during the influenza epidemic of 1987-88.with an insignificant increase in abs enteeism during the outbreak (0 .034 hours absent per hour paid versus 0 .032prior to the outbreak, 0.027 after the outbreak.and 0.025 in all other periods of that year) .Also, while the absenteeism rate among non high risk employees during the 1987-88 influenza season was higher than during the same calendar period in the previous year, the difference was not statistically significant.The vaccinat ed group: An analysis was conducted to assess if the 48 high risk employees who were vaccinated escaped the increase in absenteeism documented for the high risk departments during the 1987-88 influenza season.The data revealed no significant difference in absenteeism between outbreak and nonoutbreak seasons in this group (0.037 during the outbreak versus 0 .035prior to or 0.24 post outbreak) .Although the numbers were too small to allow firm conclusions The model assumes that 10% of employees were vaccinated (63% power to detect a 0.01 drop in absenteeism ra te, alpha=0.05,one-tailed), it is noteworthy that the rate of a bsenteeism among those vaccinated was lower during the 1987-88 influenza season (0.037) than for the corresponding p eriod in the previous year (0.046) when no influenza outbreak was documented.It was found that vaccinated h ealth care workers did not differ significantly from others in their departments by gender (95% versus 97 .1% female), marital status (62.5% versus 55.4% marri ed).number of children at home (0.9 versus 1.3), or history of relevant chronic diseases, although the vaccinated group was slightly older (mean age 3 7 .2vers u s 34.2 years, P=0 .044).Be n efits and c osts: Estimates of benefits and cost per vaccinated employee (based on the 1987-88 influenza season data) are shown in Ta ble 3. Employees in high risk departments had a 4.9% absenteeism rate during the 1987-88 influenza period (14 weeks) compared with 3. 7% throughout the rest of the 1987-88 year, and 3.8% during the corresponding 14 week period in 1986-87 when there was no influenza outbreak.Thus it was estimated that absenteeism could have been reduced by approximately 1% (4.9% to 3.8%) during the 1987-88 influenza period had influenza bee n complete ly prevented.At a vaccine efficacy of 70%, each employee vaccinated would b e expected to reduce his or her sick time by 0.70% (1 %x0.70) of the total time worked during the 14 week 'flu period'.At 38.75 h/week per employee, the estimated time of sickness saved per vaccinated employee is 38.75 h/week x 14 weeks x 0.007 (absenteeism reduction) = 3.8 h .At an average pay value of$15.72/h(for a general duly nurse II with three years of seniority) , this re presents an average expected benefit of$59.70 (3.8 h x $15.72/h) per vaccinated employee.
Direct costs associated with vaccine administration included costs of materials fo r each vaccination, ($2.25 for the vaccine and $0.25 for the needle, syringe , swabs, etc) totaling $2.50, as well as an estimated 20 m ins of the occupational health nurse's time for each vaccinated employee (including preparation , travel within the hospital, charting and filing).representing a cost of $5.87 ( 1 / s h x $17.60/h) per vaccinated employee.The only fixed annual cost of the program considered was that of the two promotion visits by the occupational health nurse to the 32 high risk wards.The estimated total lime spent to promote the program was 45 mins per ward, or 24 h total.At $17.60/h, the cost of the occupational nurse's time was $422.00, or $5.27 per vaccinated employee.
At a n estimated 20 mins lost from work duties for th e recipient for the vaccination process (travel from work assignment, completion of brief health questionnaire, interview.informed consent and vaccination).the cost component of vaccine administration was estimated at $5.25 (1/3 h x $15 .716/h)per vaccinated employee.Of 412 employees receiving the vaccine during 1984-87, 32 reported adverse reactions resulting in a total of 22 days off work Using these data, 0.05 (22/412) days were lost following vaccination.This is the same post vaccination absenteeism rate recently reported from a hospital employee influenza vaccination program in Vancouver (14) .Thus the lost time attributable to adverse vaccine reaction is 0.013 days (0.05-0 .037baseline days in the pre-influenza period) resulting in a cost of $1.58 (0.013 days x 7.75 h x $15 .716/h)per vaccinated employee.Cost-benefit analysis : As shown in Table 3, at an acceptance rate of 10%, the benefits and costs per vaccinated employee are $59 .70 and $20.47.respectively, for a net benefit of $39.23 (appendix 1 for cost-ben efit formula).If a ll 800 employees in the d esignated high risk areas were given the influenza vaccine (ie, 100% acceptance) with no change in promotion costs, the net.benefit per vaccinated employee would b e $43.98, yielding a total benefit of $35 , 182.59.
Sensitivity analyses of the variable cost components show that net savings per vaccinated employee could fall from $39.23 to $9.38 if only one-half of the excess influenza season absenteeism were prevented by vaccination or if absenteeism reduction of this magnitude occurred every other year (Table 4) .Alternatively. it would rise to $47.75 if vaccine efficacy were increased from 70  The variables wh ich impact most on the costbenefit analysis are: estim ated absenteeism due to in fl uenza, the efficacy of th e vaccine, estimated absenteeism due to adverse reactions to the vaccine, acceptance rate, and promotion costs.
A three-way sensitivity analysis shows that if lime lost due to a dverse reactions is as low as 0 .013days, net positive ben efits can be achieved at al l but the lowest estimates of rates of infl uenza absenteeism and vaccine efficacy (Table 5).Even if lime lost due to adverse reactions were as h igh as 0.02 days per vaccinated employee (20% of vaccinations, one day per adverse vaccination) , ne t positive benefits could be achieved with vaccine efficacy in the range of 60 to 80% if the increase in absenteeism rate du e to influenza is 1.00%.Another sensitivity analysis indicates that even large increases in the cost of promotion wou ld be worthwhile, if improved acceptance could be achieved (Table 6).Thus.if spending 10 times the rate ($4220) on promotion resulted in a fivefold increase in acceptance (from 10 to 50%). the net benefit per vaccinated employee would fall from $39.23 to $33 .95,but the total net benefit would rise from $3138.00 to $15,690.50, because a larger number of employees would be vaccinated .

DISCUSSION
This study was conducted to provide information which would assist in the formulation of policy for influenza vaccination of hospital employees.and specifically to determine whether vaccination of health care workers at high risk of acquiring influenza because of the type of patients they work with and the extent of exposure to these patients.will result in net savings to hospitals by reducing costs associated with employee absenteeism.In the absence of a controlled trial to demonstrate the effectiveness of reducing hospital employee absenteeism by influenza vaccination, the potential benefits can only be estimated indirectly -as in this study -using retrospective data.The magnitude of the increase in absenteeism during the 1987-88 influenza season compared to prior to the 'flu outbreak, after the outbreak.or at other limes of the year, was approximately 35%.While the vaccinated group was small , the absenteeism rate for vaccinated health care workers working in high risk areas was considerably lower than the overall rate in their departments.The estimates of the reduction in absenteeism among vaccinated employees are most likely lower than what could be expected during an epidemic for which there was a better match of vaccine to actual influenza strains.
Estimates from available data indicated a $39.23 net benefit per vaccinated employee in the influenza epidemic year studied.If all 800 employees in the designated high risk areas had been vaccinated (with no change in program costs).a total net benefit of $35,182.59 could have been realized that year.Because the value of 'unused' sick time is not returned to the employee, and if replacement nurses are hired to staff the wards affected by absenteeism, the calculated net benefit value could represent a true cost saving.
Aside from th e potential benefit to the hospital and the community, the reduced absenteeism suggested by these data likely indicates reduced morbidity for vaccinated employees.Despite the high prevalence of minor vaccine side effects reported elsewhere ( 14). this result should still represent a net benefit to the employee.
Moreover, a h ealth care worker vaccination program holds potential for reduced nosocomial in- Influenza vaccination for health care workers fluenza in hospital patients.(This latter benefit may be a more important consideration than the potential cost savings from employee absenteeism alone, and vaccination of health care workers could be justified even if the vaccination program had a net cost.) Employees without direct patient contact (ie.working in non h igh risk areas) had a slightly elevated absenteeism rate during the 1987-88 influenza season.However, it was not significantly higher than during the pre-influenza period .and was only a borderline increase in comparison with the post influenza period and other periods of the year.This finding suggests that prevention of influenza infection among hospital staff can be made more efficient by selective targeting of preventive strategies towards high risk health care workers who, in any case, are more important to target with respect to decreasing potential for nosocomial spread.
In the future, targeting of preventive strategies among hospital staff (vaccine and/ or antiviral drugs) may also be more selective if timely surveillance programs could indicate whether a virulent subtype is likely to be circulating in the community that year.For example, retrospective observations of ongoing surveillance in the United States indicate that since 1968-69, the highest excess mortality due to pneumonia and influ enza has been associated with influenza A/H3N2 viruses (15).suggesting increased virulence.The observation of excess staff absenteeism in the authors' institution in the influenza seasons of 1980-81 (13) and in 1987-88 (present study).when influenza A/H3N2 viruses were predominant, is consistent with this concept, a lthough influenza B was a lso detected in the !alter influenza period.It is also noted that the efficacy of the vaccine, influenced by the degree of match between influenza strain and vaccine, affects the cost-benefit of the program.For the present, hospitals should plan for annual vaccination programs, as influenza epidemics and virulence cannot be predicted with accuracy.
Despite the availability of free influenza vaccine, an average of less than 10% of health care workers in high risk areas accepted vaccine in any of the years of the program; at most only 41% were repeat vaccinees, and never more than 11% of those vaccinated received vaccine in more than two consecutive years.The data indicate both low acceptance rates and high attrition rates for the current program.The manner in which the program is delivered and, particularly, the extent of its promotion, clearly play a role.Schiefele and colleagues (14) reported an acceptance rate greater than 50% among hospital employees.

TABLE 1
Number of employees vaccinated and percentage acceptance of influenza vaccine in high risk and non high risk departments 1984-88 Absenteeism data were analyzed us ing a randomized b lock design analysis of variance to determine the significance of variation in rates.Demograph ic profil e data were analyzed us ing x 2 and t tests as appropriate.
CAN J INFECT DIS VOL 2 No 3 AUTUMN 1991 1985 w en ~ :::> z r---~ r---~ n ipt:lM:H JASO N DJFM AMJ lijf:tll:t:l J A SONDJFM A MJ MONTH F igur e 1) Influe nza isolates by month July 1986 to J une 19 88 with d e s ignated flu• and comparis on s tudy pe riods.Stippled bars Isolate s of irifluenza A: Solid bars Isolate s of influenza Bemployees.Absenteeism data were th er efore collected on 48 va ccinated high risk h ealth care workers for both the 1986-87 s eason ('comparis on year') and for the 'flu ' and 'non-flu' periods in the 1987-88 season.Their demographic a nd risk factor profiles were compared with a ra ndom sample of oth er e mployees in th eir departments who d id not receive the vaccine (n= 139).

TABLE 2
Table1s h ows the number and proportion of total employees who received vaccine in high risk and non h igh risk departments in each year of the program .Results show that acceptance rates were low (6 to 11 % per year) and showed no consistent trend.The highest proportion of vaccine acceptance in any ward was 43% , which occurred in 1988 on one of the geriatric nursing wards.Further analyses revealed that the same individuals did not consistently receive vaccine each year.For example, of the 146 employees vaccinated in 1986, 120 were first time vaccinees: only 26 of those vaccinated in 1986 were a lso vaccinated in 1987.Ofthe 92 vaccinated in 1987.64 were first time vaccinees, 23 were second timers.three were third timers and only two had Absenteeism rates for employees in high risk departments

TABLE 3
Best estimate of benefits and costs per vaccinated employee

TABLE 4
Univariate sensitivity analyses of major benefits and cost determinants