Reply to “Introducing a neurology department at a local hospital”

Dear Editor,

I very much appreciate the comments made regarding “Impact of introducing neurology into a local hospital in Andalusia”, based on which I can offer a better explanation for certain points in the article.

It is true that I could not analyse baseline NIHSS because the study was retrospective and NIHSS score was not recorded for patients treated by the internal medicine department. While completing a prospective study to record these data would have increased the statistical validity, it would probably have been unethical to do so. Given that NIHSS scores were missing, we measured stroke severity using the motor deficit described in patients’ medical histories at admission. Patient progress was evaluated using the modified Rankin scale, a scale typically used for this purpose in clinical trials.

I agree that analysing patient comorbidities and their influence on morbidity/mortality could have provided additional relevant information, and that this lack may constitute a bias because study groups are not necessarily homogeneous. However, we did adjust results for patients’ ages in the multivariate study (older age is probably associated with higher comorbidity). This step serves to control the confounding variable, given that the beneficial effect of treatment in a neurology department is independent from age.

Since the effect of age on patient mortality was not significant, we did not analyse the potential interaction between age and neurological care. However, it is possible that the benefits of neurological care may be more or less accentuated in older or younger patients, which may be a topic for further analysis.

The logistic regression model involved the simultaneous introduction of all variables. The model diagnosis was performed with a goodness-of-fit study (Nagelkerke R2), as this was the best model.

The ORs for mean stay could not be calculated since the dependent variable is quantitative. The ORs of the effect of neurological care on mortality and dependency were provided as decimals <1 precisely because they referred to a protective factor and not a risk factor. For an OR approaching 0, the protective factor is higher; 0.04 and 0.14 are neither minimum nor negligible values, but rather the contrary. These figures may be easily transformed into values >1 that are easier to interpret by dividing 1 by each OR, giving us values of 25 and 7 respectively. We can therefore express the data as follows: neurological care after admission reduces both hospital and one-year mortality by a factor of 25 and also reduces dependency at one year by a factor of 7.

Results are therefore robust and difficult to refute; the beneficial effect of neurological care on patient mortality and dependency is even greater when we adjust results for the other study variables.