The incremental benefits and costs of a rapid influenza A virus infection diagnostic service were studied in nursing homes in Calgary, Canada, during a single influenza season. The service was used to test 159 patients with suspected infection in a group of “experimental” nursing homes and results were compared with those for a group of “control” homes. An equal number of cases of influenza were identified in each group. Twenty-eight patients (17.6%) had confirmed cases, and 63 patients (39.6%) had probable cases. A rapid viral test result was provided much faster for patients in the experimental homes (P = .005). Both groups had the same median attack rate for influenza A virus. In experimental homes, the duration of the outbreak was shorter (P = .03), and the cost of laboratory testing and the total cost (less the hospital cost) tended to be lower (P < .2). The rapid testing service also tended to lower the overall use of resources.
Most of deaths related to influenza infections occur in people aged >65 years [1–3], and residents of long-term care facilities are particularly vulnerable to influenza outbreaks [4, 5]. Until recently, the main thrust of the influenza A virus infection control program in the Calgary Health Region (CHR), Canada, was vaccination of at-risk individuals [6, 7]. However, major outbreaks may still occur in elderly populations, despite excellent compliance with an influenza A virus vaccine program [8–10]. Additional influenza outbreak prevention measures that were introduced in CHR nursing homes coincident with this study included the grouping of patients with suspected infection into cohorts and the administration of prophylaxis with amantadine to all residents once an index case has been confirmed by a laboratory . Prompt institution of these measures has been shown to decrease the extent of respiratory influenza outbreaks in long-term care facilities, as well as the morbidity and mortality due to runaway infections [4–6].
We evaluated the use of a rapid viral diagnostic service for influenza A in conjunction with the previously outlined regional influenza control measures to determine clinical and resource outcomes for the diagnostic service. Limited data exist that assess outcome measures for such a service [11, 12]. We hypothesized that this change in laboratory service would improve clinical and cost outcomes for nursing homes to which it was provided.
Patients and Methods
Patient population and study design. The Conjoint Health Research Ethics Board at the University of Calgary approved this study. During the 1998–1999 influenza season, 12 nursing homes (of 24 in the CHR) with a total of 1705 residents were included in the study. We first characterized each nursing home according to the total number of beds and the annual rate of influenza vaccination, then matched the facilities in pairs. For each matched pair, one nursing home was randomly assigned (by coin toss) to the group that would have access to the rapid influenza A virus testing service established by Calgary Laboratory Services (CLS; i.e., the experimental group), and the other nursing home was assigned to the group that would continue to be served by the reference laboratory (i.e., the control group).
Patients with clinically suspected influenza were identified by use of the standard case definition of the Laboratory Center for Disease Control, Canada . Nasopharyngeal samples were obtained for viral testing from all patients with suspected cases of influenza in the experimental homes. A clinical case form was completed at the initial visit that included the following information: severity of illness, underlying disease or diseases, all medications taken before the onset of illness, all other laboratory tests and radiographs performed, and influenza A virus test results.
Laboratory setting and methods. CLS is a centralized regional laboratory that processes >2000 specimens a day; it operates 24 h per day and 7 days per week and provides most microbiology services for an urban health care region with a population of almost 1 million. Nurses in experimental homes were trained to obtain nasopharyngeal samples by inserting 2 intertwined polyester-tipped swabs and vigorously swabbing the nasopharynx. Both swabs were immediately placed into a single vial of Flex Trans Viral and Chlamydial Transport Medium (Bartels). All study samples were immediately transported to CLS and analyzed within 2 h of receipt. CLS analyzed all nasopharyngeal samples by use of the Directigen Flu-A assay (Becton Dickinson), which is an EIA that detects influenza A virus nucleoprotein . The assay was performed according to the manufacturer's instructions. Results were immediately relayed by telephone to the patient's ward at the experimental home.
Before this study, testing of nursing home residents for influenza A virus was performed by a reference laboratory, mainly for surveillance purposes. Both nasopharyngeal and throat swabs were tested by the reference laboratory. Nasopharyngeal samples were obtained with use of a single swab and placed into virus transport medium. Samples for viral testing were picked up daily, Monday through Friday, and rapid viral analysis and processing for viral culture were only performed the same day if the sample was received by 4 P.M. On Saturday and Sunday, rapid viral testing was only performed on an urgent basis, and routine samples obtained late Friday or during the weekend were refrigerated until they could be picked up and delivered the following Monday. The reference laboratory analyzed specimens from control nursing homes both by use of a rapid direct immunofluorescence viral assay (Dako Diagnostics), performed according to the manufacturer's instructions, and by use of standard viral culture methods . Results of routine tests were not relayed by telephone to the nursing home.
Assessment of clinical outcomes. We followed the clinical courses of patients who were confirmed to have influenza A virus infection. Infection prevention measures were documented, including isolation precautions and institution of antiviral treatment or prophylaxis. Information on antibiotic therapy was recorded, including the start and stop date for each drug and the dose and route of administration. Admission to the hospital was documented for patients who required transfer because of complications secondary to influenza A virus infection (i.e., bacterial pneumonia or other pulmonary or cardiac problems). Practitioners within the CHR Acute Care Infection Prevention and Control Program observed hospitalized patients throughout the hospital stay and completed a separate hospitalization follow-up study sheet that recorded the information outlined above. The date and cause of death were recorded for patients who died.
Assessment of influenza attack rates. For each nursing home, the initial influenza attack rate was calculated at the time that the facility identified an increased incidence of influenza-like illness above endemic rates, and the final attack rate was calculated at the time of culmination of influenza-like illness in the facility. These calculations were done using the formula previously used by Leonardi et al. : (number of cases of influenza-like illness/patient census) × 100. Rates were also estimated for the projected attack rate and the number of influenza infections prevented in each nursing home. The projected attack rate was calculated as the sum of initial attack rate and the number of new cases in the next 72 h. The percentage of preventable influenza cases was calculated as the difference between final attack rate and the projected attack rate. These hypothetical rates were then compared with the actual numbers of laboratory-confirmed cases of influenza A virus infection that occurred in each facility.
Assessment of infection control measures. We compared experimental and control nursing homes with respect to both the use of respiratory isolation procedures (i.e., grouping of patients into cohorts) attributable to the presence of possible influenza cases within each facility and with respect to the number of amantadine prescriptions that were provided. Any variations found in the use of these procedures between facilities were attributed to the influenza attack rate and whether the facility had access to rapid viral test results (i.e., attributable to the time from identification of a possible clinical case and the reporting of an influenza A virus test result).
Assessment of morbidity and mortality. Morbidity was measured as the rate of secondary complications attributed to documented influenza A illness and compared between experimental and control nursing home residents. Secondary complications included the following clinical diagnoses: secondary bacterial pneumonia; primary viral pneumonia; neurologic complications, including Guillain-Barré syndrome, encephalitis, and transverse myelitis; congestive heart failure or myocardial infarction; and respiratory failure requiring use of mechanical ventilation. Among patients with influenza A virus infection, the time from the onset of illness to recovery was also compared. Antibiotic prescription rates were compared, including the number of times the drug was prescribed, the doses prescribed, and the duration of treatment. The rate of hospitalization for residents in experimental nursing homes was compared with the rate for residents in control nursing homes with respect to specific principal diagnoses (and associated International Classification of Diseases, Ninth revision [ICD-9] codes) that have been documented to cause severe illnesses in elderly patients during influenza epidemics . The length of hospital stay was documented for patients with confirmed influenza A virus infection and correlated with the reported ICD-9 diagnosis at discharge. The mortality rate attributable to influenza A virus infection was also calculated and compared between experimental and control sites.
Assessment of resource outcomes. Health care costs associated with virologic tests were recorded. Only direct patient care costs and overhead costs for the CHR, the CLS, and the provincial Ministry of Health are considered. Cost data were obtained from numerous sources and were linked for individuals by use of each patient's personal health number. In the experimental group, the cost of performing the Directigen Flu-A test (i.e., CAN$24.18 per test) was used in the analyses and compared with the cost of the service provided by the reference laboratory. The laboratory test cost includes costs for overhead, materials and supplies, and labor.
The costs of all administered drugs associated with influenza A virus infection, including vaccines (i.e., influenza A virus vaccine and Pneumovax-23 [Merck Frosst]), antibiotics, and other related medications, were calculated as CHR cost per drug (which is specific to dose and route of administration). CHR costs were also used for other diagnostic tests, including radiography. The cost of general physician consultations was calculated according to the provincial physician fee schedule. No specialist referrals occurred in this sample. Nursing home managers were surveyed to estimate increases in facility costs due to the influenza outbreaks in each nursing home. The cost of extra consumable items (i.e., laundry, gloves, and other supplies) and staffing was divided by the number of residents in the home to calculate an incremental facility cost per subject that resulted from a given outbreak.
The personal health number of each patient hospitalized due to influenza A virus infection was used to search in the CHR computerized administrative databases during the study period to identify admissions and emergency visits at the 3 CHR adult acute care hospitals. This data was then cross-referenced with patient chart information. Inpatient costs derived from the administrative database (with use of a weighted per diem rate calculated by the CHR Finance Department) were determined for each hospitalization, and a hospital-specific per diem department cost was determined for emergency visits. The cost of capital items was not included in the analyses, because it did not vary between the new and the reference laboratory service. Results are presented both for the total cost (including hospital costs) and for the total cost less hospital costs, because a small number of hospital admissions could skew the analyses. Although the number of hospitalizations in the 2 groups was not significantly different, exclusion of hospital costs would skew the results toward the experimental group for this sample, so both sets of costs (i.e. with and without hospitalizations) are included.
Statistical analyses. Data were entered into Microsoft Access, version 6.0 (Microsoft) and analyzed by SPSS version 8.0 for Windows (SPSS). Data for a subset of 20% of cases were entered twice and checked to ensure accuracy. Medians were compared between matched nursing homes in the 2 groups by use of the Wilcoxon rank-sum test for related samples. Group proportions were compared by use of the χ2 test or Fisher's exact test for expected values of <5. Standard methods were used to calculate the sensitivity, specificity, and positive and negative predictive values of each of the test methods. For all analyses, P < .05 was considered statistically significant.
Because each nursing home had a wide range of costs, facilities could not be distinguished within the experimental and control groups. Costs were therefore compared by combining all of the costs in the 2 nursing home groups and looking at the differences in mean costs.
Patient characteristics and clinical symptoms. A total of 12 nursing homes with 1705 residents were included in the study, and 159 patients were evaluated for suspected influenza. Of these, 91 patients (57.2%) were identified as having either probable or confirmed influenza A virus infection. Sixty-three patients (39.6%) had probable infection (i.e., they met the case definition but did not have a positive influenza A virus test result). All patients with confirmed infection (28 [17.6%]) had ⩾1 influenza A virus test with positive results. As shown in table 1, an equal number of patients with suspected influenza were identified in the experimental home group and in the control nursing home group. The average age of patients in both groups was similar, although the overall age range of patients in the control nursing homes was greater. An equal number of women and men were included in both groups. The 2 : 1 ratio of women to men among patients with suspected influenza reflects the sex ratio in the nursing homes in the CHR.
Table 1 also shows the clinical symptoms manifested by patients with suspected cases of influenza. Patients from experimental and from control homes had similar symptoms at enrollment. Most of the patients had fever, and 2 patients had a temperature as high as 40°C. Respiratory tract symptoms, particularly a dry or productive cough and associated lethargy, were also reported by most patients. Less frequently manifested symptoms included sore throat, coryza, myalgia, and headache.
Laboratory confirmation of cases. Enrolled patients at experimental nursing homes were much more likely to have a diagnostic test ordered for influenza A virus. Overall, 79 patients (98.8%) with suspected cases from the experimental home group had influenza A virus testing performed, compared with only 22 patients (28.8%) in the control group (P < .0001). The median turnaround time for an influenza A virus test result for patients in the experimental nursing homes was 1 day (range, 1–2 days) versus 5 days (range, 1–20 days) for patients in the control nursing homes (P = .005; Wilcoxon rank sum test).
Clinical outcomes assessment. Clinical outcomes and data on the use of key intervention strategies, hospitalization, and mortality are outlined in table 2. Similar numbers of patients from both the experimental and control homes received an annual influenza vaccination before the start of the study. More patients in the experimental homes also received Pneumovax-23 than did patients in the control homes. Patients in control nursing homes were more likely to receive prophylaxis with amantadine than were patients in the experimental homes. Nurses in the experimental homes were less likely to begin empiric amantadine prophylaxis if the rapid influenza A virus test result was negative. The number of patients in the experimental homes who were hospitalized was somewhat higher than the number in the control homes, although this difference was not significant. Two of the patients in the experimental homes who were hospitalized were confirmed to have influenza A virus infection despite having been vaccinated. The overall mortality rate was identical in both groups. Six (43.0%) of 14 patients with probable influenza died, but the majority of these 6 patients (5 [83.3%]) had received influenza A virus vaccine. The most frequent complications attributed to acute influenza A virus infection were pulmonary problems, particularly secondary pneumonia, and exacerbation of an underlying cardiac condition (e.g., congestive heart failure).
Although the median influenza A virus attack rates were similar in the 2 groups (154 per 1000 residents [for experimental nursing homes] vs. 199 per 1000 residents [for control nursing homes]; P = .48), the duration of the influenza outbreak in control nursing homes was significantly longer: the median duration of the influenza outbreak for all nursing homes in the experimental group was 9 days, compared with 16 days in control nursing homes (P = .03). The rates for all other clinical outcomes were not significantly different between the 2 groups. The respective rates in the experimental and in the control nursing homes were as follows: prophylactic amantadine use, 447 per 1000 vs. 775 per 1000, respectively (P = .31); use of antibiotics, 683 per 1000 vs. 613 per 1000, respectively (P = .18); use of other prescription drugs, 371 per 1000 vs. 333 per 1000, respectively (P = .94); hospitalization, 123 per 1000 vs. 77 per 1000, respectively (P = 1.00); and respiratory isolation of patients with suspected influenza A virus infection, 226 per 1000 vs. 259 per 1000, respectively (P = 1.00).
Resource outcomes assessment. A comparison of the costs for each specific resource item in the experimental and the control nursing homes, as well as the total cost, is presented in table 3. A significant difference in cost was not observed for any of the resource items, although there seemed to be a trend in the experimental group toward a lower laboratory cost, and thus a lower total cost, if hospital costs are not included. However, when the hospital costs are factored in, there is a trend toward a higher total cost for the experimental group. Overall, use of Directigen Flu-A testing results in a cost per patient of CAN$29.03 less than use of current practice; use of such testing could save the health care system CAN$11,612 per year in resource costs (calculated on the basis of 400 nursing home patients), if there were no increase in the rate of hospitalization (and thus hospital costs). Conversely, if there is an underlying difference in the rate of hospitalization between the control and experimental groups, and if this true difference followed the trend exhibited in our patient sample, it could be that, despite lower laboratory costs, any savings would be offset by higher hospital costs. Because only 8.2% of the patients in our sample were hospitalized, generalizations are risky; we suggest that further study is warranted, particularly with regard to the impact of the cost of resources related to hospitalization.
Our study shows that the new laboratory service provided for the experimental nursing homes significantly diminished the overall duration of outbreaks of influenza A virus infection in nursing homes in the region (i.e., it has a higher clinical benefit than does current practice). Although previous studies have reported the clinical outcomes of prevention measures for influenza A virus infection (i.e., annual vaccination rates for high-risk populations)  and the economic provisions of these programs , few studies such as ours have been reported . Even though the rapid influenza A virus test method (i.e., direct fluorescent antibody testing) was also used by the existing reference laboratory service for the control nursing homes, test results were reported significantly faster by the new laboratory service. This occurred because nasopharyngeal samples were promptly transported to the laboratory, and a rapid viral testing result was communicated by telephone to the nursing station within 2 h of receipt of the sample. Also, a much larger number of patients with suspected cases of influenza A virus infection in the experimental homes had a diagnostic test performed, because nursing staff were properly trained to collect nasopharyngeal samples. Studies have shown that the sensitivity of specific rapid viral testing methods was optimal when nasopharyngeal samples were tested, rather than pharyngeal or throat samples [14, 16, 17].
Overall, there was also a trend toward lower laboratory costs, although further study is required to determine whether these savings are offset by an increase in the costs associated with hospitalization. By extending a study such as this beyond a single influenza season, trends observed for our sample could be further investigated. Although it is possible that more resources might be expended in providing a highly efficient rapid viral diagnostic service, a decision would have to be made as to whether the improved benefits observed in this sample are offset by any increases in costs.
The timely deployment of antiviral prophylaxis depends on the availability of the influenza A virus test result. Because amantadine has significant side effects in elderly patients, one of the newer but more costly antiviral agents may replace this drug as the main prophylactic agent for influenza A virus infection [18–20]. Zanamivir is cleared slowly from the respiratory tract after being inhaled, which makes it an attractive option for prophylaxis against influenza A virus infection . Recent clinical efficacy studies, however, have shown that this drug is most effective when prescribed during the first 1–2 days of the illness [21, 22]. Zanamivir and other new anti–influenza A virus agents may be inappropriately prescribed in nursing homes if a rapid viral testing service, as described here, is not available.
Although this study supports the implementation of a more efficient rapid influenza A virus diagnostic service for nursing homes in Calgary, it may not be possible to generalize its findings to other health care jurisdictions that do not have a centralized microbiology service similar to that described here. In addition, costs in our region may not be applicable in other locations. Further study is required to fully address questions of the impact on resources of providing a rapid viral testing service in place of standard laboratory service, particularly as related to the issue of hospitalization.