How to pivot wider a table with tidyverse if some factor levels are empty

I want to reshape a dataframe from long to wide format. The code is:

structure(list(periodo = c(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 
1, 1, 1, 1, 1, 1), lugar = c("A", "A", "A", "A", "B", "B", "B", 
"B", "B", "B", "A", "A", "A", "B", "B", "B", "B", "B"), nihss = structure(c(2L, 
3L, 4L, 5L, 1L, 2L, 3L, 4L, 5L, 6L, 3L, 4L, 5L, 2L, 3L, 4L, 5L, 
6L), levels = c("0", "1-4", "5-15", "16-20", "20+", "9999"), class = "factor"), 
    porcent = c(5, 56, 17, 22, 1, 12, 53, 15, 18, 1, 30, 30, 
    40, 6, 37, 27, 28, 2)), row.names = c(NA, -18L), class = "data.frame") 

One column, nihss, is a factor variable, with ordered levels:

stroke$nihss <- factor(stroke$nihss, levels = c("0","1-4","5-15","16-20","20+","9999"))

Some levels are empty for some combinations of the named variables, periodo and lugar, in pivot_wider.
My code to pivot these data is:

stroke %>% 
  pivot_wider(names_from = c(lugar, periodo), values_from = porcent, values_fill = 0) %>% 
  kable()

The code works as expected, and the reshaped table has four new columns. However, the nihss' levels are not ordered according to the original order. The first ("0") and the last ("9999") appears after the remaining categories:

nihss	A_0	B_0	A_1	B_1
1-4	5	12	0	6
5-15	56	53	30	37
16-20	17	15	30	27
20+	22	18	40	28
0	0	1	0	0
9999	0	1	0	2

I expect to get a pivoted table, where the levels of the factor are shown as originally were ordered. That is:

"0","1-4","5-15","16-20","20+","9999"

Add dplyr::arrange() to your output, like:

[...]

 stroke$nihss <- factor(stroke$nihss, levels = c("0","1-4","5-15","16-20","20+","9999"))

 stroke |>
   tidyr::pivot_wider(names_from = c(lugar, periodo), values_from = porcent, values_fill = 0) |>
   dplyr::arrange(nihss) |>
   knitr::kable() 

^{Created on 2023-06-29 with reprex v2.0.2}