Impact on health and workdays
The model results suggest a substantial impact on general practitioners, even with very conservative assumptions. For the "more severe" scenario a mortality rate of 65 per 100 000 is predicted (albeit for just one pandemic wave). This is much less than the total population rate for the 1918 pandemic in New Zealand of 745 per 100 000 , but it is more than United States total population rates for the 1957 Asian flu pandemic (22 per 100 000) and the 1968 Hong Kong flu pandemic (14 per 100 000) .
The results suggest that the major contributor to lost workdays will be episodes of uncomplicated illness that do not require hospitalisation. If time spent caring for sick relatives is considered (i.e., as in the "more severe" scenario) then this also made a substantial contribution to the total workdays lost. The impact of lost workdays will be magnified by the increased demand on the medical workforce, as has recently been modelled for primary care consultations and hospitalisations in New Zealand , and for critical care services in both New Zealand and Australia .
Implications for the health sector
There are several broad strategies to reduce the impact of an influenza pandemic on health care workers. First, infection control strategies aimed at doctors need to be in place. These measures include basic hygiene practices and also mask use may be appropriate (depending on risk ). Health authorities and doctors themselves could also stockpile and then use antivirals at the appropriate time. Such stockpiling has already commenced at a national level in New Zealand and various other countries . Recent modelling work indicates that access to enough antivirals could substantially reduce the number of clinical cases and hospitalisations in the population .
Second, pandemic planning needs to include specific measures to maintain the functional capacity of health care workers, bearing in mind that the impact of an influenza pandemic is likely to vary between urban and rural areas. While exposure to infection may be less in relatively isolated rural areas, such areas generally have far less "spare" health care capacity, should GPs be incapacitated. General practices and health authorities can consider plans to provide care for the ill dependents of their medical staff so as to reduce absenteeism rates. Through other pandemic planning activities they can also potentially reduce the overall impact of a pandemic and hence demands on their staff. For example, rapid action at the start of the pandemic to cancel elective procedures could enhance workforce capacity. Establishing dedicated primary care assessment centres for patients with suspected influenza could also reduce overall GP workload.
Third, strategies are needed to manage the psychological impact of pandemic influenza on health care workers. Surveys of such workers show that they report a lower willingness to report for duty for infectious diseases epidemics (SARS, smallpox) than for most other forms of catastrophic disasters (environmental disasters, mass casualty incidents) . Experience with SARS also demonstrated the psychological importance of having well-designed policies and protocols in place. Even in situations where health care workers perceive themselves to be at increased risk, they report feeling reassured by simple protective measures based on sound epidemiological principles, when implemented in a timely manner . A review of the foundations for a SARS preparedness and response plan has specifically highlighted the importance of both appropriate staffing and support .
Fourth, improving health sector surge capacity now would be desirable as the New Zealand health sector is often running at stretched capacity (e.g., especially emergency departments ). Expanding existing services such as the "Healthline" (a free telephone information service to the public staffed by nurses) may also be worthwhile. Similarly, active promotion of key websites with information on managing influenza (e.g., as per the CDC website ) could be publicised each winter season. All such measures would benefit the public prior to a pandemic as well as potentially reducing the demands on the medical workforce in the primary care and secondary care settings during a pandemic.
Finally, a greater focus on the primary care nursing workforce would be of benefit. Following the implementation of the Primary Health Care Strategy there has been a rapid shift to capitation funding of general practices, and an attendant increased focus on team-based primary care (principally GPs and practice nurses). This trend raises the possibility of increasing substitution of GP work roles by nurses. This type of substitution has occurred for a decade or more in a range of community-governed non-profit practices and other capitation-funded practices [29, 30]. A recent review of the medical workforce in New Zealand also highlights the potential efficiencies from some role shifting from doctors to other health workers . Expanding such a non-medical health workforce, while also vulnerable to the infection during a pandemic, would provide a buffer for the GP workforce in the event of attrition of GP capacity.
Limitations with the modelling
The uncertainties associated with pandemic influenza mean that estimating its future impact is problematic. This model could substantially underestimate the true impact because the new strain may be particularly infectious and/or virulent, and the incidence rate for clinical illness might be higher for doctors given their likely occupational risk . For example, one review of nosocomial outbreaks reported a health care worker incidence rate to be as high as 60% . Furthermore, doctors may be relatively slow to seek care for themselves – especially at the time of a national crisis when their professional obligations are greatest. Other parameters used in the modelling may also have been overly conservative, such as the extent of the healthy worker effect among doctors and the amount of time off work taken to care for sick relatives (which was zero in the baseline model and fairly small at 0.5 days in the "more severe" scenario). There was also no consideration in the model of absenteeism effects from fear of infection (e.g., in the case of particularly virulent strains). Indeed, this absenteeism effect could be more important than actual disease in reducing health sector capacity.
Although we consider that the baseline results are more likely to underestimate than to overestimate the impact of a future influenza pandemic, there are still plausible reasons why they could be overestimates. These include the following:
various international and national public health interventions (as recommended by WHO ) may reduce the impact of pandemic influenza;
at least for subsequent pandemic waves, an appropriate vaccine may be available;
antivirals could prevent infection and reduce morbidity amongst the medical workforce and the rest of the population ;
improved treatment could lower hospitalisation and mortality rates (relative to the figures used in this model).
This modelling could be further refined to address some of the limitations detailed above. Clarifying the prevalence of "high-risk" conditions among the medical workforce would be a particularly important refinement along with improving the estimates of time off work to care for relatives (or even absenteeism from fear of infection). Expanding such modelling to other parts of the health sector workforce is also desirable, along with exploring the extent that such research is generalisable to other threats (e.g., from other new emerging infectious threats, including those from bioterrorism).