Our study found that the NH quality as measured by ADL decline, weight loss, pressure ulcers, and infections are significantly and negatively impacted by influenza. This finding was reproduced consistently in each of our analyses. As expected, our control measures of nursing home quality, including persistent pain and restraint use, were largely unaffected by influenza severity, nor did they correlate with influenza mortality. In a typical year, with approximately 1.23 million long-stay NH residents, using the average increases in influenza severity and mortality during the influenza season (quarters 1, 2, and 4) relative to the summer months (quarter 3) of lowest flu activity, our model estimated that an additional 12,284 long-stay NH residents experience an ADL decline of four points or larger; an additional 15,168 NH residents experience significant weight loss; an additional 6,284 NH residents experience new or worsening pressure ulcers; and an additional 29,753 NH residents experience an infection.
Taken together, these findings suggest that seasonal influenza has a major negative impact on the quality of life for the surviving NH elderly population. The ADL decline, weight loss, worsening pressure ulcers—presumed secondary to immobility, and increased rates of infections reflect an advanced level of disability and functional dependence. This likely results in increased personnel and health-care expenditures for the NH facilities.32
The association between influenza and ADL decline has been reported previously in a case-control study among frail NH residents from a limited geographical location, over 2 consecutive influenza seasons.33
Our study confirms these findings at the NH facility level, using a 6-year longitudinal analysis of a large-sample size, over the entire country. The scope of the effects we find on ADL decline and the other outcomes of interest provide insight into the strength of the forces at work over which even the best nursing facilities have little control. Influenza season, particularly in a city hard hit by influenza, will increase the work load on staff as they struggle to keep residents mobile, their skin from breaking down as they are sick in bed, and their food intake at reasonable levels. However, in the case of restraints, persistent pain, antipsychotic use, and infection rates, there appear to be NH specific unobserved factors that explain a large portion the variance (above 50%) of these outcomes beyond the explanation provided by seasonal effects of influenza. This indicates that these outcomes are potentially more amenable to reductions through increased staffing or other quality improvement measures than outcomes like large ADL decline, weight loss, or pressure ulcers, whose variance appears better explained by seasonal influenza. This suggests that research is needed on how high vs. less high quality homes cope with exposure to influenza.
Our findings have policy implications for the Center for Medicare/Medicaid Services current public reporting model, which posts facilities’ quality measures using a 5-star scale based in part on ADL decline and pressure ulcer prevalence, since only the facility’s most recent measure is posted. Those facilities in cities or states in which influenza was particularly severe will necessarily have worse performance on these measures. 34, 35
This indicates the importance and need for seasonal adjustment for these quality measures. For persons considering NHs in different cities or states, proper seasonal adjustment would provide a better comparative measure of NH quality. More indirectly, given that our functional decline outcomes are risk factors for hospitalizations of NH residents, and gven the introduction by CMS payment penalties for hospitals which have high risk-adjusted re-hospitalization rates, hospitals in cities/regions that have more severe influenza seasons could be unfairly penalized relative to those in regions with more benign influenza seasons. Careful examination of the effect of seasonal variation on re-hospitalizations from nursing homes (and from the community) is certainly warranted.36
Our study has several limitations that deserve mention. First, influenza mortality is measured indirectly, as P& I deaths, as opposed to laboratory-confirmed influenza, and as such, limits sensitivity and specificity. Other viruses that circulate in the winter, notably respiratory syncytial virus, can cause clinical symptoms indistinguishable from influenza, and their mortality can be mistakenly attributed to influenza, 1
reducing the specificity of our influenza mortality variable. On the other hand, influenza-related deaths from causes other than pulmonary, such as cardiovascular events, for example, are not included in our variable, limiting its sensitivity. Reassuringly, the strong relationship we observed with influenza mortality at the city level on the local functional status outcomes persisted.
Since influenza severity is the result of local and regional epidemiological surveillance based on laboratory-confirmation and clinical symptoms, its increased specificity may explain the stronger associations with the decline in functional status outcomes we found, validating our results.
A second limitation is that our analysis includes data derived from the MDS, a large, administrative, non-audited dataset that may be subject to reporting errors. However, the MDS has been shown to have high levels of inter-rater reliability particularly on the measures of ADL performance. 37
Finally, because our data on clinical outcomes have been aggregated at the NH-facility level, some of the differences present at the individual-level may have been lost in the analysis. Further studies examining the impact of influenza on functional status in NH residents using non-aggregated, large cohort individual-level data are important to confirm our findings.
Our longitudinal facility fixed effects models control for any unobserved facility level characteristics that are time invariant, or quasi time-invariant, and the use of autoregressive errors corrects standard errors for correlation over time. But there is a possibility that some time-varying confounders have not been accounted for.
A final important limitation was the imprecision of influenza vaccination as reported in the OSCAR NH surveys, which prevented us from having reliable vaccination measures for all NHs, plus lack of information on NH staff vaccination. As such, the lack of significance of influenza vaccination on functional decline outcomes is an open question that requires additional research.